diff --git a/.agents/skills/update-models/SKILL.md b/.agents/skills/update-models/SKILL.md
index 9ffb5628..fe6ff729 100644
--- a/.agents/skills/update-models/SKILL.md
+++ b/.agents/skills/update-models/SKILL.md
@@ -1,6 +1,6 @@
 ---
 name: update-models
-description: Discover latest OpenAI, Anthropic, Google/Gemini, Tzafon, and Yutori models and verify computer-use support. Use when updating CUA model defaults, checking new model releases, auditing provider-native computer tool actions, or comparing provider metadata, official examples, and smoke-test results.
+description: Discover latest OpenAI, Anthropic, Google/Gemini, Meta, Tzafon, and Yutori models and verify computer-use support. Use when updating CUA model defaults, checking new model releases, auditing provider-native computer tool actions, or comparing provider metadata, official examples, and smoke-test results.
 ---
 
 # Update Models
@@ -9,14 +9,14 @@ Use this workflow to keep CUA current with provider model releases and computer-
 
 ## Quick Start
 
-1. Verify credentials are available: `OPENAI_API_KEY`, `ANTHROPIC_API_KEY`, `GOOGLE_API_KEY` or `GEMINI_API_KEY`, `TZAFON_API_KEY`, and `YUTORI_API_KEY`.
+1. Verify credentials are available: `OPENAI_API_KEY`, `ANTHROPIC_API_KEY`, `GOOGLE_API_KEY` or `GEMINI_API_KEY`, `META_API_KEY`, `TZAFON_API_KEY`, and `YUTORI_API_KEY`.
 2. If credentials live in `~/AGENTS.md`, load them into the current shell without printing them:
 
 ```bash
 eval "$(python3 - <<'PY'
 import pathlib, re, shlex
 text = pathlib.Path('~/AGENTS.md').expanduser().read_text()
-for key in ['OPENAI_API_KEY', 'ANTHROPIC_API_KEY', 'GOOGLE_API_KEY', 'TZAFON_API_KEY', 'YUTORI_API_KEY']:
+for key in ['OPENAI_API_KEY', 'ANTHROPIC_API_KEY', 'GOOGLE_API_KEY', 'META_API_KEY', 'TZAFON_API_KEY', 'YUTORI_API_KEY']:
     m = re.search(r'export\s+' + re.escape(key) + r'=(?:"([^"]+)"|([^\s\n]+))', text)
     if m:
         print(f'export {key}={shlex.quote(m.group(1) or m.group(2))}')
@@ -68,6 +68,15 @@ When live discovery finds a new model with passing smoke tests, update `packages
 
 ## Provider Checks
 
+Meta:
+
+- Discover with the OpenAI SDK against `https://api.meta.ai/v1` using `META_API_KEY`.
+- Smoke-test the Responses API with screenshot input and explicit function tools matching CUA's canonical actions.
+- Pass condition: response output contains a `function_call` for one of the supplied browser actions.
+- Use `store: true` plus `previous_response_id` for CUA tool loops. Meta rejects `include: ["reasoning.encrypted_content"]` on requests that set `previous_response_id`.
+- Set `parallel_tool_calls: false` because browser actions mutate shared state.
+- Treat Meta computer use as custom-function-tool support, not a provider-native `{ type: "computer" }` tool.
+
 OpenAI:
 
 - Discover with `OpenAI().models.list()` and optionally `models.retrieve(modelId)`.
@@ -116,6 +125,7 @@ Yutori:
 Run action probes when updating adapters or when docs/examples show drift:
 
 ```bash
+npx tsx .agents/skills/update-models/reference/discover-models.ts --provider meta --models muse-spark-1.1
 npx tsx .agents/skills/update-models/reference/native-action-probe.ts --provider openai --model gpt-5.5
 npx tsx .agents/skills/update-models/reference/native-action-probe.ts --provider anthropic --model claude-opus-4-7
 npx tsx .agents/skills/update-models/reference/native-action-probe.ts --provider gemini --model gemini-3-flash-preview
@@ -160,6 +170,7 @@ All CUA model and adapter support lives in `packages/ai` (`@onkernel/cua-ai`). W
   - Update the snapshot in `packages/ai/docs/supported-models.md` to match.
 
 - New provider-native action, response field, or tool version:
+  - Meta: update `packages/ai/src/providers/meta/index.ts` and `provider.ts`, including Responses threading and reasoning compatibility.
   - OpenAI: update `packages/ai/src/providers/openai/index.ts` and its action vocabulary, plus the shared canonical types in `packages/ai/src/providers/common.ts` if the action set changes.
   - Anthropic: update the `ANTHROPIC_CUA_ACTION_TYPES` set in `packages/ai/src/providers/anthropic/actions.ts` and `index.ts`. The computer tool version and `computer-use-*` beta header are selected by `pi-ai` per model, so a new dated tool version usually means bumping `@earendil-works/pi-ai`, not editing this package.
   - Gemini: update `packages/ai/src/providers/gemini/index.ts`, including coordinate handling if needed.
diff --git a/.agents/skills/update-models/reference/README.md b/.agents/skills/update-models/reference/README.md
index 4569bda7..686fa65f 100644
--- a/.agents/skills/update-models/reference/README.md
+++ b/.agents/skills/update-models/reference/README.md
@@ -11,10 +11,11 @@ These scripts support the `update-models` skill. Run them from the repository ro
   - `OPENAI_API_KEY`
   - `ANTHROPIC_API_KEY`
   - `GOOGLE_API_KEY` or `GEMINI_API_KEY`
+  - `META_API_KEY`
   - `TZAFON_API_KEY`
   - `YUTORI_API_KEY`
 
-The scripts never print API keys. Smoke tests are non-destructive: they ask each model to emit a provider-native computer-use tool call, then inspect the response without executing the action.
+The scripts never print API keys. Smoke tests are non-destructive: they ask each model to emit a computer-use tool call, then inspect the response without executing the action. Meta uses supplied function tools; other providers may use provider-native computer tools.
 
 ## Common Commands
 
diff --git a/.agents/skills/update-models/reference/audit-official-examples.ts b/.agents/skills/update-models/reference/audit-official-examples.ts
index 3940e017..247d295a 100644
--- a/.agents/skills/update-models/reference/audit-official-examples.ts
+++ b/.agents/skills/update-models/reference/audit-official-examples.ts
@@ -5,7 +5,7 @@ import { basename, join, resolve } from "node:path";
 import { spawnSync } from "node:child_process";
 import process from "node:process";
 
-type Provider = "openai" | "anthropic" | "gemini" | "yutori";
+type Provider = "openai" | "anthropic" | "gemini" | "meta" | "yutori";
 
 interface ExampleRepo {
 	provider: Provider;
@@ -51,6 +51,14 @@ const EXAMPLES: ExampleRepo[] = [
 		confidence: "provider-owned",
 		patterns: ["computer_use", "ComputerUse", "function_call", "functionCall", "FunctionResponse", "safety_decision"],
 	},
+	{
+		provider: "meta",
+		name: "meta-model-cookbook",
+		repo: "https://github.com/meta-models/meta-model-cookbook.git",
+		confidence: "provider-owned",
+		pathHint: "03_use_cases/13_macos_cua",
+		patterns: ["muse-spark-1.1", "function_call", "function_call_output", "previous_response_id", "reasoning.encrypted_content", "parallel_tool_calls"],
+	},
 	{
 		provider: "yutori",
 		name: "kernel-cli-yutori-template",
@@ -65,6 +73,7 @@ const ACTION_REGEXES: Record<Provider, RegExp[]> = {
 	openai: [/\b(click|double_click|scroll|type|wait|keypress|drag|move|screenshot)\b/g],
 	anthropic: [/\b(screenshot|left_click|right_click|middle_click|double_click|triple_click|left_click_drag|mouse_move|key|type|scroll|hold_key|wait|left_mouse_down|left_mouse_up|cursor_position|zoom)\b/g],
 	gemini: [/\b(open_web_browser|open_web|wait_5_seconds|go_back|go_forward|search|navigate|click_at|hover_at|type_text_at|key_combination|scroll_document|scroll_at|drag_and_drop)\b/g],
+	meta: [/\b(screenshot|left_click|right_click|middle_click|double_click|triple_click|left_click_drag|mouse_move|key|type|scroll|hold_key|wait|left_mouse_down|left_mouse_up)\b/g],
 	yutori: [/\b(left_click|double_click|triple_click|right_click|scroll|type|key_press|hover|drag|wait|refresh|go_back|goto_url|mouse_move|middle_click|mouse_down|mouse_up|go_forward|hold_key|extract_elements|find|set_element_value|execute_js)\b/g],
 };
 
diff --git a/.agents/skills/update-models/reference/discover-models.ts b/.agents/skills/update-models/reference/discover-models.ts
index c8bb4db5..8984a138 100644
--- a/.agents/skills/update-models/reference/discover-models.ts
+++ b/.agents/skills/update-models/reference/discover-models.ts
@@ -1,11 +1,12 @@
 #!/usr/bin/env tsx
+import { existsSync } from "node:fs";
 import { readFile, writeFile } from "node:fs/promises";
 import { homedir } from "node:os";
 import { join } from "node:path";
 import process from "node:process";
 import { parse as parseToml } from "smol-toml";
 
-type Provider = "openai" | "anthropic" | "gemini" | "tzafon" | "yutori";
+type Provider = "openai" | "anthropic" | "gemini" | "meta" | "tzafon" | "yutori";
 
 interface Args {
 	provider: Provider | "all";
@@ -38,7 +39,7 @@ interface ModelResult {
 	cua?: Record<string, unknown>;
 }
 
-const PROVIDERS: Provider[] = ["openai", "anthropic", "gemini", "tzafon", "yutori"];
+const PROVIDERS: Provider[] = ["openai", "anthropic", "gemini", "meta", "tzafon", "yutori"];
 const TZAFON_KNOWN_MODELS = [
 	"tzafon.northstar-cua-fast",
 ];
@@ -109,7 +110,7 @@ function usage(): never {
   npx tsx .agents/skills/update-models/reference/discover-models.ts --provider openai --models gpt-5.5,gpt-5.4
 
 Options:
-  --provider <all|openai|anthropic|gemini|tzafon|yutori>
+  --provider <all|openai|anthropic|gemini|meta|tzafon|yutori>
   --models <comma-separated model ids>    Smoke-test explicit models instead of inferred candidates.
   --candidate-limit <n>                  Max inferred candidates per provider. Default: 20.
   --no-smoke                             Only list metadata.
@@ -137,6 +138,7 @@ async function runProvider(provider: Provider, args: Args): Promise<Record<strin
 		if (provider === "openai") return await discoverOpenAI(args);
 		if (provider === "anthropic") return await discoverAnthropic(args);
 		if (provider === "gemini") return await discoverGemini(args);
+		if (provider === "meta") return await discoverMeta(args);
 		if (provider === "tzafon") return await discoverTzafon(args);
 		if (provider === "yutori") return await discoverYutori(args);
 		throw new Error(`unknown provider ${provider satisfies never}`);
@@ -175,6 +177,79 @@ async function discoverOpenAI(args: Args): Promise<Record<string, unknown>> {
 	return { provider: "openai", metadata_source: "client.models.list()", models, candidates };
 }
 
+async function discoverMeta(args: Args): Promise<Record<string, unknown>> {
+	const OpenAI = await importDefault("openai", "OpenAI");
+	const apiKey = process.env.META_API_KEY;
+	const client = new OpenAI({ apiKey, baseURL: "https://api.meta.ai/v1" });
+	const rawModels = await collectAsync(client.models.list());
+	const models: ModelResult[] = rawModels.map((m) => ({
+		id: String(m.id),
+		display_name: String(m.id),
+		created_at: typeof m.created === "number" && m.created > 0 ? new Date(m.created * 1000).toISOString() : null,
+		raw: m,
+		supports_generation: true,
+	}));
+	const candidates = explicitOrCandidates(args, models.map((model) => model.id));
+	if (args.smoke) {
+		await Promise.all(candidates.map(async (id) => {
+			const model = models.find((candidate) => candidate.id === id) ?? { id, display_name: id, supports_generation: true };
+			model.computer_use = await smokeMeta(client, id);
+			if (!models.find((candidate) => candidate.id === id)) models.unshift(model);
+		}));
+	}
+	await annotateCuaSupport("meta", models);
+	return { provider: "meta", metadata_source: "Meta Model API models.list()", models, candidates };
+}
+
+async function smokeMeta(client: any, model: string): Promise<SmokeResult> {
+	try {
+		const screenshotPath = [
+			join(process.cwd(), "examples", "screenshot.png"),
+			join(process.cwd(), "packages", "ai", "examples", "screenshot.png"),
+		].find(existsSync);
+		if (!screenshotPath) throw new Error("could not find packages/ai/examples/screenshot.png");
+		const screenshot = await readFile(screenshotPath);
+		const response = await client.responses.create({
+			model,
+			store: false,
+			parallel_tool_calls: false,
+			max_output_tokens: 512,
+			reasoning: { effort: "low" },
+			input: [{
+				role: "user",
+				content: [
+					{ type: "input_text", text: "Call the click tool for the sign in link. Do not answer only in text." },
+					{ type: "input_image", image_url: `data:image/png;base64,${screenshot.toString("base64")}` },
+				],
+			}],
+			tools: [{
+				type: "function",
+				name: "click",
+				description: "Click at normalized 0-1000 screen coordinates.",
+				parameters: {
+					type: "object",
+					properties: { x: { type: "number" }, y: { type: "number" } },
+					required: ["x", "y"],
+					additionalProperties: false,
+				},
+			}],
+		});
+		const output: any[] = response.output ?? [];
+		const calls = output.filter((item) => item?.type === "function_call");
+		return {
+			status: calls.length > 0 ? "pass" : "inconclusive",
+			tool_name: "function_tools",
+			tool_version: null,
+			beta_header: null,
+			observed_actions: unique(calls.map((call) => call?.name).filter(Boolean)),
+			response_item_types: unique(output.map((item) => item?.type).filter(Boolean)),
+			error: null,
+		};
+	} catch (err) {
+		return smokeError(err, { tool_name: "function_tools" });
+	}
+}
+
 function likelyOpenAIGenerationModel(id: string): boolean {
 	const lower = id.toLowerCase();
 	if (OPENAI_EXCLUDE.some((needle) => lower.includes(needle))) return false;
diff --git a/.agents/skills/update-models/reference/provider-doc-drift.ts b/.agents/skills/update-models/reference/provider-doc-drift.ts
index c60fbd42..bc652685 100644
--- a/.agents/skills/update-models/reference/provider-doc-drift.ts
+++ b/.agents/skills/update-models/reference/provider-doc-drift.ts
@@ -2,7 +2,7 @@
 import { readFile, writeFile } from "node:fs/promises";
 import process from "node:process";
 
-type Provider = "openai" | "anthropic" | "gemini" | "yutori";
+type Provider = "openai" | "anthropic" | "gemini" | "meta" | "yutori";
 
 interface Args {
 	examples: string;
@@ -30,6 +30,11 @@ const DOCS: Record<Provider, string[]> = {
 		"https://ai.google.dev/gemini-api/docs/computer-use",
 		"https://ai.google.dev/api/models",
 	],
+	meta: [
+		"https://dev.meta.ai/docs/getting-started/cookbook/computer-use-macos",
+		"https://dev.meta.ai/docs/features/responses",
+		"https://dev.meta.ai/docs/features/tool-calling",
+	],
 	yutori: [
 		"https://docs.yutori.com/reference/navigator",
 		"https://docs.yutori.com/reference/n1",
@@ -42,6 +47,7 @@ const LOCAL_FILES: Record<Provider, string> = {
 	openai: "packages/ai/src/providers/openai/index.ts",
 	anthropic: "packages/ai/src/providers/anthropic/actions.ts",
 	gemini: "packages/ai/src/providers/gemini/index.ts",
+	meta: "packages/ai/src/providers/meta/index.ts",
 	yutori: "packages/ai/src/providers/yutori/actions.ts",
 };
 
@@ -49,6 +55,7 @@ const ACTION_REGEXES: Record<Provider, RegExp> = {
 	openai: /\b(click|double_click|scroll|type|wait|keypress|drag|move|screenshot)\b/g,
 	anthropic: /\b(screenshot|left_click|right_click|middle_click|double_click|triple_click|left_click_drag|mouse_move|key|type|scroll|hold_key|wait|left_mouse_down|left_mouse_up|cursor_position|zoom)\b/g,
 	gemini: /\b(open_web_browser|open_web|wait_5_seconds|go_back|go_forward|search|navigate|click_at|hover_at|type_text_at|key_combination|scroll_document|scroll_at|drag_and_drop)\b/g,
+	meta: /\b(click|double_click|mouse_down|mouse_up|scroll|type|wait|keypress|drag|move|screenshot|goto|back|forward|url|cursor_position|left_click|right_click|middle_click|triple_click|left_click_drag|mouse_move|key|hold_key|left_mouse_down|left_mouse_up)\b/g,
 	yutori: /\b(left_click|double_click|triple_click|right_click|scroll|type|key_press|hover|drag|wait|refresh|go_back|go_forward|goto_url|mouse_move|middle_click|mouse_down|mouse_up|hold_key|extract_elements|find|set_element_value|execute_js)\b/g,
 };
 
@@ -123,8 +130,8 @@ async function checkProvider(provider: Provider, examples: any): Promise<Record<
 		documented_actions: sorted(documentedActions),
 		example_repo_actions: sorted(exampleActions),
 		repo_supported_actions: sorted(localActions),
-		unknown_documented_actions: sorted(difference(new Set(documentedActions), new Set(localActions))),
-		unknown_example_actions: sorted(difference(new Set(exampleActions), new Set(localActions))),
+		unknown_documented_actions: provider === "meta" ? [] : sorted(difference(new Set(documentedActions), new Set(localActions))),
+		unknown_example_actions: provider === "meta" ? [] : sorted(difference(new Set(exampleActions), new Set(localActions))),
 		response_fields_from_examples: sorted(example.response_fields ?? []),
 		notes: notesFor(provider, documentedToolVersions, exampleToolVersions, localToolVersions),
 	};
@@ -154,6 +161,9 @@ function notesFor(provider: Provider, documentedToolVersions: string[], exampleT
 	if (provider === "gemini") {
 		notes.push("Gemini official computer use emits predefined function-call names; keep this separate from CUA custom function declarations.");
 	}
+	if (provider === "meta") {
+		notes.push("Meta uses developer-defined function tools, so provider action-name drift does not apply; compare payload threading and coordinate conventions instead.");
+	}
 	if (provider === "yutori") {
 		notes.push("Yutori Navigator emits OpenAI-compatible tool_calls for built-in browser actions; local AgentTools should execute those names but outbound payloads should not duplicate the built-in browser schemas.");
 		notes.push("Track n1 vs n1.5 separately because n1.5 can add tool_set/disable_tools and expanded browser actions.");
diff --git a/.agents/skills/update-models/reference/report-schema.md b/.agents/skills/update-models/reference/report-schema.md
index 7247f449..52e9c93e 100644
--- a/.agents/skills/update-models/reference/report-schema.md
+++ b/.agents/skills/update-models/reference/report-schema.md
@@ -12,6 +12,7 @@ Use this shape for JSON reports and the same fields when writing a Markdown summ
     "openai": {},
     "anthropic": {},
     "gemini": {},
+    "meta": {},
     "tzafon": {},
     "yutori": {}
   },
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 1aac9651..4f78ed44 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -132,6 +132,7 @@ jobs:
           OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
           ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
           GOOGLE_API_KEY: ${{ secrets.GOOGLE_API_KEY }}
+          META_API_KEY: ${{ secrets.META_API_KEY }}
           TZAFON_API_KEY: ${{ secrets.TZAFON_API_KEY }}
           YUTORI_API_KEY: ${{ secrets.YUTORI_API_KEY }}
         run: npm run test:integration --workspace @onkernel/cua-ai
@@ -158,6 +159,7 @@ jobs:
           OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
           ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
           GOOGLE_API_KEY: ${{ secrets.GOOGLE_API_KEY }}
+          META_API_KEY: ${{ secrets.META_API_KEY }}
           TZAFON_API_KEY: ${{ secrets.TZAFON_API_KEY }}
           YUTORI_API_KEY: ${{ secrets.YUTORI_API_KEY }}
         run: npm test --workspace @onkernel/cua-agent -- test/e2e.live.test.ts
diff --git a/README.md b/README.md
index dbbec9bf..97da85fd 100644
--- a/README.md
+++ b/README.md
@@ -12,7 +12,7 @@ cua "go to news.ycombinator.com and tell me the top 3 story titles"
 
 ## Why this exists
 
-Every frontier model now ships its own first-party "computer use" tool:
+Frontier models expose computer use through different tool protocols:
 
 - **OpenAI gpt-5.5**: a built-in `computer` tool that emits actions like
   `{type:"click", x, y}`, `{type:"scroll", x, y, scroll_x, scroll_y}`,
@@ -23,6 +23,8 @@ Every frontier model now ships its own first-party "computer use" tool:
 - **Google gemini-2.5-pro / gemini-3.x**: a set of predefined
   computer-use functions (`click_at`, `type_text_at`, `scroll_at`,
   `navigate`, `go_back`, …) with 0-1000 normalized coordinates.
+- **Meta Muse Spark 1.1**: OpenAI-compatible Responses API function calls
+  paired with screenshot inputs and 0-1000 normalized coordinates.
 - **Yutori Navigator n1 / n1.5**: OpenAI-compatible `chat.completions`
   responses with built-in browser action `tool_calls` like `left_click`,
   `goto_url`, `type`, and `scroll` in 0-1000 normalized coordinates.
@@ -100,6 +102,7 @@ npx tsx packages/cli/src/cli.ts --help
 export OPENAI_API_KEY=sk-...                 # for gpt-5.5
 export ANTHROPIC_API_KEY=sk-ant-...          # for claude-opus-4-7
 export GOOGLE_API_KEY=...                    # for gemini-3-flash-preview
+export META_API_KEY=...                      # for muse-spark-1.1
 export YUTORI_API_KEY=yt_...                 # for n1.5-latest
 export KERNEL_API_KEY=sk_...                 # always required
 
@@ -115,6 +118,9 @@ cua -p --model claude-opus-4-7 "Same prompt"
 # Gemini 3 Flash (built-in computer use)
 cua -p --model gemini-3-flash-preview "Same prompt"
 
+# Meta Muse Spark
+cua -p --model meta:muse-spark-1.1 "Same prompt"
+
 # Yutori Navigator
 cua -p --model n1.5-latest "Same prompt"
 
diff --git a/package-lock.json b/package-lock.json
index 663eb65b..fdcf3859 100644
--- a/package-lock.json
+++ b/package-lock.json
@@ -15,6 +15,7 @@
       ],
       "devDependencies": {
         "@types/node": "22.18.4",
+        "smol-toml": "^1.7.0",
         "tsx": "^4.21.0",
         "typescript": "5.9.3"
       },
@@ -5569,6 +5570,19 @@
       "dev": true,
       "license": "ISC"
     },
+    "node_modules/smol-toml": {
+      "version": "1.7.0",
+      "resolved": "https://registry.npmjs.org/smol-toml/-/smol-toml-1.7.0.tgz",
+      "integrity": "sha512-aqVvWoyO21L23mb+drl4RmMXbf6N7FdHjAhTRA9ZBL7apWBgfWC16KjrASI+1p9GAroljyMHj6fK67i0UiTNvQ==",
+      "dev": true,
+      "license": "BSD-3-Clause",
+      "engines": {
+        "node": ">= 18"
+      },
+      "funding": {
+        "url": "https://github.com/sponsors/cyyynthia"
+      }
+    },
     "node_modules/source-map-js": {
       "version": "1.2.1",
       "resolved": "https://registry.npmjs.org/source-map-js/-/source-map-js-1.2.1.tgz",
diff --git a/package.json b/package.json
index 1f5397e0..0e3e7a67 100644
--- a/package.json
+++ b/package.json
@@ -22,6 +22,7 @@
   },
   "devDependencies": {
     "@types/node": "22.18.4",
+    "smol-toml": "^1.7.0",
     "tsx": "^4.21.0",
     "typescript": "5.9.3"
   }
diff --git a/packages/agent/README.md b/packages/agent/README.md
index 7b7404cd..77d3dfed 100644
--- a/packages/agent/README.md
+++ b/packages/agent/README.md
@@ -127,7 +127,7 @@ Both classes mirror pi constructor shapes and behavior, with minimal additions:
   live browser session
 - `toolResultImageReplayLimit` to control how many recent tool-result images
   are included in each model request
-- `responseThreading` to choose whether OpenAI and Tzafon continue from
+- `responseThreading` to choose whether OpenAI, Meta, and Tzafon continue from
   provider-stored response state or receive the full request context each time
 
 ### Context management
@@ -177,7 +177,7 @@ pi provides several context-management layers with different lifetimes:
   It is broader than the image limit and is not called automatically by this
   package.
 
-`responseThreading` defaults to `true`. For OpenAI and Tzafon Responses APIs,
+`responseThreading` defaults to `true`. For OpenAI, Meta, and Tzafon Responses APIs,
 that means each request can continue from `previous_response_id`; the provider
 keeps earlier response state and CUA sends only messages added since that
 response. Because the provider retains that state, removing an older image from
@@ -192,6 +192,7 @@ If auth callbacks are omitted, both classes default to CUA env var conventions:
 - OpenAI: `OPENAI_API_KEY`
 - Anthropic: `ANTHROPIC_OAUTH_TOKEN` or `ANTHROPIC_API_KEY`
 - Gemini: `GOOGLE_API_KEY` or `GEMINI_API_KEY`
+- Meta: `META_API_KEY`
 - Tzafon: `TZAFON_API_KEY`
 - Yutori: `YUTORI_API_KEY`
 
@@ -223,7 +224,7 @@ browser session does carry over. No screenshot is returned automatically;
 request one on a follow-up turn when the model needs to see the page.
 Playwright-level failures come back as tool content (so the model can adapt)
 rather than thrown errors. Verified e2e
-against Anthropic, Tzafon, and Yutori CUA models; OpenAI and Google are
+against Anthropic, Meta, Tzafon, and Yutori CUA models; OpenAI and Google are
 unit-tested.
 
 ### Model Switching
diff --git a/packages/agent/src/agent.ts b/packages/agent/src/agent.ts
index ce62089f..9d57eaca 100644
--- a/packages/agent/src/agent.ts
+++ b/packages/agent/src/agent.ts
@@ -156,7 +156,9 @@ export type CuaAgentOptions = Omit<AgentOptions, "initialState"> & {
 	emptyResponseRecovery?: CuaEmptyResponseRecoveryOptions;
 	/** Maximum tool-result images included from message history per provider request. Defaults to 4; false disables projection. */
 	toolResultImageReplayLimit?: ToolResultImageReplayLimit;
-	/** Chain OpenAI and Tzafon requests through provider-stored response state. Defaults to true. */
+	/**
+	 * Chain OpenAI, Meta, and Tzafon requests through provider-stored response state. Defaults to true.
+	 */
 	responseThreading?: boolean;
 	/** Optional CUA-level retries around each provider request. Disabled by default. */
 	retry?: CuaRetryOptions;
@@ -202,7 +204,9 @@ export type CuaAgentHarnessOptions<
 	emptyResponseRecovery?: CuaEmptyResponseRecoveryOptions;
 	/** Maximum tool-result images included from message history per provider request. Defaults to 4; false disables projection. */
 	toolResultImageReplayLimit?: ToolResultImageReplayLimit;
-	/** Chain OpenAI and Tzafon requests through provider-stored response state. Defaults to true. */
+	/**
+	 * Chain OpenAI, Meta, and Tzafon requests through provider-stored response state. Defaults to true.
+	 */
 	responseThreading?: boolean;
 	/** Optional CUA-level retries around each provider request. Disabled by default. */
 	retry?: CuaRetryOptions;
diff --git a/packages/agent/test/e2e.live.test.ts b/packages/agent/test/e2e.live.test.ts
index 41e3436a..55fbc1d6 100644
--- a/packages/agent/test/e2e.live.test.ts
+++ b/packages/agent/test/e2e.live.test.ts
@@ -20,6 +20,7 @@ type ProviderCase = {
 		| "openai:gpt-5.5"
 		| "anthropic:claude-opus-4-7"
 		| "google:gemini-3-flash-preview"
+		| "meta:muse-spark-1.1"
 		| "tzafon:tzafon.northstar-cua-fast"
 		| "yutori:n1.5-latest";
 	prompt: string;
@@ -79,6 +80,19 @@ const cases: ProviderCase[] = [
 		timeoutMs: 300_000,
 		ciOptInEnvVar: "CUA_E2E_GEMINI",
 	},
+	{
+		name: "meta",
+		apiKeyEnvVar: "META_API_KEY",
+		modelRef: "meta:muse-spark-1.1",
+		prompt: [
+			"Use the tool named `screenshot` exactly once to inspect the browser.",
+			"Pass empty arguments (`{}`).",
+			"Do not call any other tools.",
+			"After the tool result, provide a one-sentence summary.",
+		].join("\n"),
+		expectToolCalls: true,
+		timeoutMs: 180_000,
+	},
 	{
 		name: "tzafon",
 		apiKeyEnvVar: "TZAFON_API_KEY",
@@ -125,11 +139,15 @@ type RunStats = {
 	assistantErrors: string[];
 };
 
+function apiKeyForCase(c: ProviderCase): string | undefined {
+	return process.env[c.apiKeyEnvVar];
+}
+
 function shouldRunCase(c: ProviderCase): boolean {
 	if (!LIVE) return false;
 	if (!KERNEL_API_KEY) return false;
 	if (c.ciOptInEnvVar && process.env.CI && process.env[c.ciOptInEnvVar] !== "1") return false;
-	return Boolean(process.env[c.apiKeyEnvVar]);
+	return Boolean(apiKeyForCase(c));
 }
 
 function shouldRunSwitchCase(c: ModelSwitchCase): boolean {
@@ -229,7 +247,7 @@ describe("Cua live e2e", () => {
 					const agent = new CuaAgent({
 						browser,
 						client,
-						getApiKey: () => process.env[c.apiKeyEnvVar],
+						getApiKey: () => apiKeyForCase(c),
 						afterToolCall: async () => ({ terminate: true }),
 						initialState: {
 							model: c.modelRef,
@@ -257,7 +275,7 @@ describe("Cua live e2e", () => {
 						client,
 						model: c.modelRef,
 						getApiKeyAndHeaders: async () => {
-							const apiKey = process.env[c.apiKeyEnvVar];
+							const apiKey = apiKeyForCase(c);
 							return apiKey ? { apiKey } : undefined;
 						},
 					});
@@ -287,8 +305,8 @@ describe("Cua live e2e", () => {
 						browser,
 						client,
 						getApiKey: (provider) => {
-							if (provider === c.from.modelRef.split(":")[0]) return process.env[c.from.apiKeyEnvVar];
-							if (provider === c.to.modelRef.split(":")[0]) return process.env[c.to.apiKeyEnvVar];
+							if (provider === c.from.modelRef.split(":")[0]) return apiKeyForCase(c.from);
+							if (provider === c.to.modelRef.split(":")[0]) return apiKeyForCase(c.to);
 							return undefined;
 						},
 						initialState: {
@@ -323,11 +341,11 @@ describe("Cua live e2e", () => {
 						model: c.from.modelRef,
 						getApiKeyAndHeaders: async (model) => {
 							if (model.provider === c.from.modelRef.split(":")[0]) {
-								const apiKey = process.env[c.from.apiKeyEnvVar];
+								const apiKey = apiKeyForCase(c.from);
 								return apiKey ? { apiKey } : undefined;
 							}
 							if (model.provider === c.to.modelRef.split(":")[0]) {
-								const apiKey = process.env[c.to.apiKeyEnvVar];
+								const apiKey = apiKeyForCase(c.to);
 								return apiKey ? { apiKey } : undefined;
 							}
 							return undefined;
diff --git a/packages/ai/README.md b/packages/ai/README.md
index 97ab4b14..f8c9739c 100644
--- a/packages/ai/README.md
+++ b/packages/ai/README.md
@@ -20,6 +20,7 @@ check these environment variables, in order:
 | `openai`    | `OPENAI_API_KEY`                            |
 | `anthropic` | `ANTHROPIC_OAUTH_TOKEN`, `ANTHROPIC_API_KEY` |
 | `google`    | `GOOGLE_API_KEY`, `GEMINI_API_KEY`          |
+| `meta`      | `META_API_KEY`                              |
 | `tzafon`    | `TZAFON_API_KEY`                            |
 | `yutori`    | `YUTORI_API_KEY`                            |
 
@@ -210,6 +211,7 @@ provider/tool metadata.
 getCuaModel("openai:gpt-5.5");
 getCuaModel("anthropic:claude-opus-4-7");
 getCuaModel("google:gemini-3-flash-preview");
+getCuaModel("meta:muse-spark-1.1");
 getCuaModel("tzafon:tzafon.northstar-cua-fast");
 getCuaModel("yutori:n1.5-latest");
 ```
@@ -226,7 +228,7 @@ list of CUA-supporting models per provider.
 `CuaProvider` is the string union of provider IDs this package targets:
 
 ```ts
-type CuaProvider = "openai" | "anthropic" | "google" | "tzafon" | "yutori";
+type CuaProvider = "openai" | "anthropic" | "google" | "meta" | "tzafon" | "yutori";
 ```
 
 The IDs match pi-ai's `Model.provider` values exactly. `providerForModel(model)`
@@ -332,7 +334,7 @@ definitions and executors; it is forwarded to the provider module's
   with pi's builtin providers plus CUA's adjustments: OpenAI models route
   through cua's `openai-cua-responses` stream (threads
   `previous_response_id`), Google resolves `GOOGLE_API_KEY` or
-  `GEMINI_API_KEY`, and the Tzafon/Yutori providers are registered. Each call
+  `GEMINI_API_KEY`, and the Meta/Tzafon/Yutori providers are registered. Each call
   returns an independent collection.
 - `cuaModels(): MutableModels` — the shared default collection, created on
   first use. `CuaAgent` and `CuaAgentHarness` stream through it unless given
@@ -370,6 +372,7 @@ Per-provider canonical action subsets (each namespace exports its list as
 | `openai`    | all 16                                                                              |
 | `anthropic` | 13 — everything except `back`, `forward`, `url`; adds `computer_batch` by default  |
 | `gemini`    | all 16                                                                              |
+| `meta`      | all 16                                                                              |
 | `tzafon`    | all 16 (replaced on the wire by Tzafon's native `computer_use` tool)                |
 | `yutori`    | 13 — everything except `screenshot`, `url`, `cursor_position` (local mirrors only)  |
 
@@ -404,6 +407,7 @@ Current coordinate contracts:
 - `openai`: pixel coordinates
 - `anthropic`: pixel coordinates, matching Anthropic's computer-use quickstart
 - `gemini`: normalized coordinates in the 0-999 range ([source](https://ai.google.dev/gemini-api/docs/computer-use))
+- `meta`: normalized coordinates in the 0-1000 range ([source](https://dev.meta.ai/docs/getting-started/cookbook/computer-use-macos))
 - `yutori`: normalized coordinates in the 0-1000 range ([source](https://docs.yutori.com/reference/navigator), [SDK helper](https://github.com/yutori-ai/yutori-sdk-python/blob/main/yutori/navigator/coordinates.py))
 - `tzafon`: normalized coordinates in the 0-999 range ([source](https://docs.lightcone.ai/guides/coordinates/), [model card](https://huggingface.co/Tzafon/Northstar-CUA-Fast))
 
@@ -480,7 +484,7 @@ type CuaNavigationInput = {
 
 ## Provider Namespaces
 
-Every provider namespace (`openai`, `anthropic`, `gemini`, `tzafon`,
+Every provider namespace (`openai`, `anthropic`, `gemini`, `meta`, `tzafon`,
 `yutori`) follows one convention:
 
 - `computerTools(options?)` and `computerToolExecutors(options?)`
@@ -504,6 +508,9 @@ Provider-specific extras:
   `computer_use_extra` navigation aliases `OPENAI_EXTRA_TOOL_NAME`,
   `OPENAI_EXTRA_TOOL_DESCRIPTION`, `OpenAIExtraSchema`, `OpenAIExtraInput`
 - `anthropic`: `ANTHROPIC_BATCH_TOOL_NAME` (`"computer_batch"`)
+- `meta`: the `meta-responses` stream adapter (`META_RESPONSES_API`,
+  `streamMetaResponses`, `streamSimpleMetaResponses`, `MetaResponsesOptions`),
+  with `previous_response_id` threading and Meta-compatible reasoning payloads
 - `tzafon`: the `tzafon-responses` stream adapter (`TZAFON_RESPONSES_API`,
   `streamTzafonResponses`, `streamSimpleTzafonResponses`,
   `TzafonResponsesOptions` with `keepToolNames`), `tzafonComputerUseOnPayload`
diff --git a/packages/ai/docs/supported-models.md b/packages/ai/docs/supported-models.md
index 1a51d8e1..0812cfff 100644
--- a/packages/ai/docs/supported-models.md
+++ b/packages/ai/docs/supported-models.md
@@ -17,7 +17,7 @@ with CUA-only entries that pi-ai does not ship yet.
 
 ## `openai`
 
-API: `openai-responses` · coordinates: pixel
+Coordinates: pixel
 
 Family matches (root + numeric revision/dated-snapshot suffixes):
 
@@ -27,7 +27,7 @@ Family matches (root + numeric revision/dated-snapshot suffixes):
 
 ## `anthropic`
 
-API: `anthropic-messages` · coordinates: pixel
+Coordinates: pixel
 
 Family matches (root + numeric revision/dated-snapshot suffixes):
 
@@ -42,7 +42,7 @@ Source: [Anthropic computer use docs](https://docs.anthropic.com/en/docs/build-w
 
 ## `google`
 
-API: `google-generative-ai` · coordinates: normalized 0–999
+Coordinates: normalized 0–999
 
 Model refs use the `google:` prefix; `gemini:` is accepted as an alias.
 
@@ -58,9 +58,22 @@ Google's native `tools.computer_use` request wrapper instead.
 
 Source: [Gemini computer use docs](https://ai.google.dev/gemini-api/docs/computer-use).
 
+## `meta`
+
+Coordinates: normalized 0–1000
+
+Exact IDs:
+
+- `muse-spark-1.1`
+
+Muse Spark uses Meta's OpenAI-compatible Responses API with ordinary function
+tools. CUA continues tool loops through `previous_response_id`.
+
+Source: [Meta computer-use cookbook](https://dev.meta.ai/docs/getting-started/cookbook/computer-use-macos).
+
 ## `tzafon`
 
-API: `tzafon-responses` · coordinates: normalized 0–999
+Coordinates: normalized 0–999
 
 Exact IDs:
 
@@ -70,7 +83,7 @@ Exact IDs:
 
 ## `yutori`
 
-API: `yutori-chat-completions` · coordinates: normalized 0–1000
+Coordinates: normalized 0–1000
 
 Exact IDs:
 
diff --git a/packages/ai/examples/quickstart.ts b/packages/ai/examples/quickstart.ts
index cf9256e0..fed43c6f 100644
--- a/packages/ai/examples/quickstart.ts
+++ b/packages/ai/examples/quickstart.ts
@@ -10,6 +10,7 @@ import {
 // Switch providers by setting CUA_MODEL (and the matching API key env var):
 //   anthropic:claude-opus-4-7        ANTHROPIC_API_KEY
 //   google:gemini-3-flash-preview    GOOGLE_API_KEY
+//   meta:muse-spark-1.1              META_API_KEY
 //   tzafon:tzafon.northstar-cua-fast TZAFON_API_KEY
 //   yutori:n1.5-latest               YUTORI_API_KEY
 const modelRef = (process.env.CUA_MODEL ?? "openai:gpt-5.5") as CuaModelRef;
diff --git a/packages/ai/src/api-keys.ts b/packages/ai/src/api-keys.ts
index cfbbed74..769599f3 100644
--- a/packages/ai/src/api-keys.ts
+++ b/packages/ai/src/api-keys.ts
@@ -12,6 +12,7 @@ const CUA_PROVIDER_API_KEY_ENV_VARS: Record<CuaProvider, readonly string[]> = {
 	openai: ["OPENAI_API_KEY"],
 	anthropic: ["ANTHROPIC_OAUTH_TOKEN", "ANTHROPIC_API_KEY"],
 	google: ["GOOGLE_API_KEY", "GEMINI_API_KEY"],
+	meta: ["META_API_KEY"],
 	tzafon: ["TZAFON_API_KEY"],
 	yutori: ["YUTORI_API_KEY"],
 };
diff --git a/packages/ai/src/index.ts b/packages/ai/src/index.ts
index e71e3e61..54e05158 100644
--- a/packages/ai/src/index.ts
+++ b/packages/ai/src/index.ts
@@ -7,6 +7,7 @@ export * from "./runtime-spec";
 export * from "./providers/common";
 export * as anthropic from "./providers/anthropic/index";
 export * as gemini from "./providers/gemini/index";
+export * as meta from "./providers/meta/index";
 export * as openai from "./providers/openai/index";
 export * as tzafon from "./providers/tzafon/index";
 export * as yutori from "./providers/yutori/index";
diff --git a/packages/ai/src/models.ts b/packages/ai/src/models.ts
index 5653e901..df574b44 100644
--- a/packages/ai/src/models.ts
+++ b/packages/ai/src/models.ts
@@ -1,9 +1,10 @@
 import type { Api, Model } from "@earendil-works/pi-ai";
 import { getBuiltinModel, getBuiltinModels } from "@earendil-works/pi-ai/providers/all";
+import { META_RESPONSES_API } from "./providers/meta/provider";
 import { OPENAI_CUA_RESPONSES_API } from "./providers/openai/provider";
 
 /** Providers with curated computer-use model support. */
-export type CuaProvider = "openai" | "anthropic" | "google" | "tzafon" | "yutori";
+export type CuaProvider = "openai" | "anthropic" | "google" | "meta" | "tzafon" | "yutori";
 
 /** Provider-qualified model reference, e.g. `"openai:gpt-5.5"` or `"google:gemini-3-flash-preview"`. */
 export type CuaModelRef = `${CuaProvider}:${string}`;
@@ -20,7 +21,7 @@ export interface CuaModelInfo {
 }
 
 /** All providers this package curates computer-use models for. */
-export const CUA_PROVIDERS: readonly CuaProvider[] = ["openai", "anthropic", "google", "tzafon", "yutori"];
+export const CUA_PROVIDERS: readonly CuaProvider[] = ["openai", "anthropic", "google", "meta", "tzafon", "yutori"];
 
 /**
  * How a {@link CuaModelAnnotation} matches model ids.
@@ -78,6 +79,9 @@ export const CUA_MODEL_ANNOTATIONS: Record<CuaProvider, readonly CuaModelAnnotat
 		// gemini-3-pro-preview is intentionally absent: Google retired it and
 		// the API now returns 404 "model no longer available".
 	],
+	meta: [
+		{ match: { kind: "exact", id: "muse-spark-1.1" }, source: "https://dev.meta.ai/docs/getting-started/cookbook/computer-use-macos" },
+	],
 	tzafon: [
 		{ match: { kind: "exact", id: "tzafon.northstar-cua-fast" }, source: "https://huggingface.co/Tzafon/Northstar-CUA-Fast" },
 		{ match: { kind: "exact", id: "tzafon.northstar-cua-fast-1.6" }, source: "https://huggingface.co/Tzafon/Northstar-CUA-Fast" },
@@ -100,6 +104,8 @@ const CUA_MODEL_OVERRIDES: Record<CuaProvider, readonly Model<Api>[]> = {
 	openai: [],
 	anthropic: [],
 	google: [],
+	// pi-ai 0.80.3 predates Meta's models.dev catalog entry.
+	meta: [cuaModel("meta", "muse-spark-1.1", "Muse Spark 1.1")],
 	tzafon: [
 		cuaModel("tzafon", "tzafon.northstar-cua-fast", "Tzafon Northstar CUA Fast"),
 		cuaModel("tzafon", "tzafon.northstar-cua-fast-1.6", "Tzafon Northstar CUA Fast 1.6"),
@@ -193,13 +199,17 @@ export function getCuaModel(ref: CuaModelRef): Model<Api> {
 	throw new Error(`CUA model "${ref}" is supported but not registered. Add it to pi-ai (models.dev) or CUA_MODEL_OVERRIDES.`);
 }
 
-// Route OpenAI CUA models to cua's own openai-cua-responses stream provider,
-// which threads previous_response_id. Registry-resolved models (gpt-5.4,
-// gpt-5.4-mini, gpt-5.5) otherwise carry pi-ai's builtin "openai-responses" api.
+// Route OpenAI-compatible CUA models to provider-specific Responses transports
+// that thread previous_response_id. Registry-resolved models otherwise carry
+// pi-ai's builtin "openai-responses" api.
 export function routeCuaApi(model: Model<Api>): Model<Api> {
-	return model.provider === "openai" && model.api !== OPENAI_CUA_RESPONSES_API
-		? { ...model, api: OPENAI_CUA_RESPONSES_API }
-		: model;
+	if (model.provider === "openai" && model.api !== OPENAI_CUA_RESPONSES_API) {
+		return { ...model, api: OPENAI_CUA_RESPONSES_API };
+	}
+	if (model.provider === "meta" && model.api !== META_RESPONSES_API) {
+		return { ...model, api: META_RESPONSES_API };
+	}
+	return model;
 }
 
 /** Return the {@link CuaProvider} for a concrete model, or throw when it is not a CUA provider. */
@@ -251,7 +261,7 @@ function cuaModel(provider: CuaProvider, id: string, name: string): Model<Api> {
 		id,
 		name,
 		provider,
-		reasoning: provider === "openai" || provider === "anthropic" || provider === "google",
+		reasoning: provider === "openai" || provider === "anthropic" || provider === "google" || provider === "meta",
 		input: ["text", "image"],
 		cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
 	} satisfies Partial<Model<Api>>;
@@ -263,6 +273,19 @@ function cuaModel(provider: CuaProvider, id: string, name: string): Model<Api> {
 			return { ...base, api: "anthropic-messages", baseUrl: "https://api.anthropic.com", contextWindow: 200_000, maxTokens: 64_000 } as Model<Api>;
 		case "google":
 			return { ...base, api: "google-generative-ai", baseUrl: "https://generativelanguage.googleapis.com/v1beta", contextWindow: 1_048_576, maxTokens: 65_536 } as Model<Api>;
+		case "meta":
+			// Meta documents the 1,048,576-token context window, and its
+			// computer-use cookbook configures 128,000 maximum output tokens.
+			return {
+				...base,
+				api: META_RESPONSES_API,
+				baseUrl: "https://api.meta.ai/v1",
+				thinkingLevelMap: { off: null, xhigh: "xhigh" },
+				cost: { input: 1.25, output: 4.25, cacheRead: 0.15, cacheWrite: 0 },
+				contextWindow: 1_048_576,
+				maxTokens: 128_000,
+				compat: { supportsDeveloperRole: true, sendSessionIdHeader: false, supportsLongCacheRetention: true },
+			} as Model<Api>;
 		case "tzafon":
 			return { ...base, api: "tzafon-responses", baseUrl: "https://api.lightcone.ai", contextWindow: 128_000, maxTokens: 4_096 } as Model<Api>;
 		case "yutori":
diff --git a/packages/ai/src/providers.ts b/packages/ai/src/providers.ts
index 9a5d1c18..cdd8b225 100644
--- a/packages/ai/src/providers.ts
+++ b/packages/ai/src/providers.ts
@@ -14,6 +14,7 @@ import { builtinModels } from "@earendil-works/pi-ai/providers/all";
 import { cuaApiKeyEnvVarsForProvider } from "./api-keys";
 import { cuaOverrideModels } from "./models";
 import { ANTHROPIC_NATIVE_API_BETA_HEADERS, withAnthropicBetaHeader } from "./providers/anthropic/native";
+import { META_RESPONSES_API, streamMetaResponses, streamSimpleMetaResponses } from "./providers/meta/provider";
 import { OPENAI_CUA_RESPONSES_API, streamOpenAIResponses, streamSimpleOpenAIResponses } from "./providers/openai/provider";
 import { streamSimpleTzafonResponses, streamTzafonResponses, TZAFON_RESPONSES_API } from "./providers/tzafon/provider";
 import { streamSimpleYutori, streamYutori, YUTORI_CHAT_COMPLETIONS_API } from "./providers/yutori/provider";
@@ -32,7 +33,7 @@ import { streamSimpleYutori, streamYutori, YUTORI_CHAT_COMPLETIONS_API } from ".
  *   `anthropic-beta` header merged in.
  * - `google` resolves its API key from `GOOGLE_API_KEY` or `GEMINI_API_KEY`
  *   (pi's builtin only reads `GEMINI_API_KEY`).
- * - `tzafon` and `yutori` are CUA-only providers pi does not ship.
+ * - `meta`, `tzafon`, and `yutori` are CUA-only providers pi does not ship.
  *
  * Each call returns an independent collection; register additional providers
  * or credentials on it freely. Use {@link cuaModels} for the shared default.
@@ -45,6 +46,7 @@ export function createCuaModels(options?: CreateModelsOptions): MutableModels {
 	if (anthropic) models.setProvider(withAnthropicNativeTools(anthropic));
 	const google = models.getProvider("google");
 	if (google) models.setProvider(withGoogleEnvKeys(google));
+	models.setProvider(metaProvider());
 	models.setProvider(tzafonProvider());
 	models.setProvider(yutoriProvider());
 	return models;
@@ -106,6 +108,17 @@ function withGoogleEnvKeys(base: Provider): Provider {
 	return { ...base, auth: { ...base.auth, apiKey: envApiKeyAuth("Google API key", cuaApiKeyEnvVarsForProvider("google")) } };
 }
 
+function metaProvider(): Provider {
+	return createProvider({
+		id: "meta",
+		name: "Meta",
+		baseUrl: "https://api.meta.ai/v1",
+		auth: { apiKey: envApiKeyAuth("Meta Model API key", cuaApiKeyEnvVarsForProvider("meta")) },
+		models: cuaOverrideModels("meta"),
+		api: { stream: streamMetaResponses, streamSimple: streamSimpleMetaResponses },
+	});
+}
+
 function tzafonProvider(): Provider {
 	return createProvider({
 		id: "tzafon",
@@ -128,6 +141,7 @@ function yutoriProvider(): Provider {
 	});
 }
 
+export { META_RESPONSES_API, streamMetaResponses, streamSimpleMetaResponses };
 export { OPENAI_CUA_RESPONSES_API, streamOpenAIResponses, streamSimpleOpenAIResponses };
 export { TZAFON_RESPONSES_API, streamSimpleTzafonResponses, streamTzafonResponses };
 export { YUTORI_CHAT_COMPLETIONS_API, streamSimpleYutori, streamYutori };
diff --git a/packages/ai/src/providers/common.ts b/packages/ai/src/providers/common.ts
index 49e89d48..dc54a620 100644
--- a/packages/ai/src/providers/common.ts
+++ b/packages/ai/src/providers/common.ts
@@ -2,10 +2,12 @@ import {
 	Type,
 	type Api,
 	type AssistantMessage,
+	type Context,
 	type Message,
 	type Model,
 	type SimpleStreamOptions,
 	type Static,
+	type StreamOptions,
 	type TSchema,
 	type Tool,
 } from "@earendil-works/pi-ai";
@@ -25,6 +27,15 @@ export * from "../native-tools";
  */
 export const CUA_ACTION_TYPES = CUA_DEFAULT_COMPUTER_ACTION_TYPES;
 
+export const DEFAULT_COMPUTER_INSTRUCTIONS = `You control a Kernel cloud browser through individual browser tools. Base each action on the latest observed state and request a screenshot when you need a fresh view.`;
+
+/**
+ * Build the provider-neutral system prompt for canonical computer tools.
+ */
+export function buildDefaultComputerSystemPrompt(opts: { suffix?: string } = {}): string {
+	return [DEFAULT_COMPUTER_INSTRUCTIONS, opts.suffix].filter(Boolean).join("\n\n");
+}
+
 type ObjectSchemaWithProperties = TSchema & { properties: Record<string, TSchema> };
 
 function createCuaActionArgumentSchema(action: CuaActionType, mode: CuaMode): TSchema {
@@ -263,7 +274,7 @@ export type CuaPayloadHook = (payload: unknown, model: Model<Api>, context?: Cua
  * pi-ai `SimpleStreamOptions` plus CUA request controls consumed by provider
  * adapters. Pass `keepToolNames` for caller tools that must survive native
  * tool-set substitution; use `disableResponseThreading` for a full-context
- * OpenAI or Tzafon request.
+ * OpenAI, Meta, or Tzafon request.
  */
 export interface CuaSimpleStreamOptions extends SimpleStreamOptions, ResponseThreadingOptions {
 	keepToolNames?: readonly string[];
@@ -275,6 +286,38 @@ export interface ResponseThreadingOptions {
 	disableResponseThreading?: boolean;
 }
 
+type ResponsesOnPayload = NonNullable<StreamOptions["onPayload"]>;
+
+/**
+ * Options shared by OpenAI-compatible Responses transports that support threading.
+ */
+export interface ResponsesThreadingOptions extends ResponseThreadingOptions {
+	onPayload?: ResponsesOnPayload;
+}
+
+/**
+ * Prepare an OpenAI-compatible Responses request using stored response state.
+ * When a prior response exists, only messages after it are sent.
+ */
+export function threadResponsesRequest(
+	context: Context,
+	api: Api,
+	options: ResponsesThreadingOptions | undefined,
+): { context: Context; onPayload: ResponsesOnPayload; previousResponseId?: string } {
+	const delta = responseThreadingEnabled(options) ? responseThreadingDelta(context.messages, api) : undefined;
+	const previousResponseId = delta?.previousResponseId;
+	const messages = previousResponseId && delta ? delta.deltaMessages : context.messages;
+	const onPayload: ResponsesOnPayload = async (payload, model) => {
+		const threaded = {
+			...(payload as Record<string, unknown>),
+			store: true,
+			...(previousResponseId ? { previous_response_id: previousResponseId } : {}),
+		};
+		return options?.onPayload ? ((await options.onPayload(threaded, model)) ?? threaded) : threaded;
+	};
+	return { context: messages === context.messages ? context : { ...context, messages }, onPayload, previousResponseId };
+}
+
 /**
  * Whether a Responses API provider should thread requests with
  * `previous_response_id` + delta input instead of sending the full message
diff --git a/packages/ai/src/providers/meta/index.ts b/packages/ai/src/providers/meta/index.ts
new file mode 100644
index 00000000..26f7822d
--- /dev/null
+++ b/packages/ai/src/providers/meta/index.ts
@@ -0,0 +1,41 @@
+import { assertComputerModeOnly, buildDefaultComputerSystemPrompt, computerToolExecutors, computerTools } from "../common";
+import type { ComputerToolCoordinateSystem, ComputerToolsOptions, CuaProviderModule } from "../common";
+
+export {
+	buildDefaultComputerSystemPrompt as buildMetaSystemPrompt,
+	CUA_ACTION_TYPES as META_CUA_ACTION_TYPES,
+	DEFAULT_COMPUTER_INSTRUCTIONS as META_COMPUTER_INSTRUCTIONS,
+	computerToolExecutors,
+	computerTools,
+	createCuaActionSchema as createActionSchema,
+} from "../common";
+export type {
+	CuaAction as MetaAction,
+	ComputerToolsOptions,
+} from "../common";
+export {
+	META_RESPONSES_API,
+	streamMetaResponses,
+	streamSimpleMetaResponses,
+	threadMetaRequest,
+} from "./provider";
+export type { MetaResponsesOptions } from "./provider";
+
+// Muse Spark reports screenshot positions on a normalized 0-1000 grid.
+// Source: https://dev.meta.ai/docs/features/image-understanding#coordinate-system
+export function coordinateSystem(): ComputerToolCoordinateSystem {
+	return { type: "normalized", range: [0, 1000] };
+}
+
+export const providerModule = {
+	toolDefinitions: (options?: ComputerToolsOptions) => {
+		assertComputerModeOnly("meta", options);
+		return computerTools(options);
+	},
+	toolExecutors: (options?: ComputerToolsOptions) => {
+		assertComputerModeOnly("meta", options);
+		return computerToolExecutors(options);
+	},
+	coordinateSystem,
+	buildSystemPrompt: buildDefaultComputerSystemPrompt,
+} satisfies CuaProviderModule;
diff --git a/packages/ai/src/providers/meta/provider.ts b/packages/ai/src/providers/meta/provider.ts
new file mode 100644
index 00000000..dffa9e15
--- /dev/null
+++ b/packages/ai/src/providers/meta/provider.ts
@@ -0,0 +1,62 @@
+import {
+	type Context,
+	type OpenAIResponsesOptions as PiOpenAIResponsesOptions,
+	type StreamFunction,
+} from "@earendil-works/pi-ai";
+import {
+	stream as piStreamOpenAIResponses,
+	streamSimple as piStreamSimpleOpenAIResponses,
+} from "@earendil-works/pi-ai/api/openai-responses";
+import {
+	type CuaSimpleStreamOptions,
+	type ResponsesThreadingOptions,
+	threadResponsesRequest,
+} from "../common";
+
+export const META_RESPONSES_API = "meta-responses";
+
+/**
+ * Stream options for Meta's OpenAI-compatible Responses API.
+ */
+export interface MetaResponsesOptions extends PiOpenAIResponsesOptions, ResponsesThreadingOptions {}
+
+/**
+ * Apply Meta's computer-use payload constraints on top of shared Responses threading.
+ */
+export function threadMetaRequest(context: Context, options: ResponsesThreadingOptions | undefined) {
+	const callerOnPayload = options?.onPayload;
+	const threaded = threadResponsesRequest(context, META_RESPONSES_API, {
+		...options,
+		onPayload: async (payload, model) => {
+			const constrained: Record<string, unknown> = { ...(payload as Record<string, unknown>), parallel_tool_calls: false };
+			delete constrained.include;
+			return callerOnPayload ? ((await callerOnPayload(constrained, model)) ?? constrained) : constrained;
+		},
+	});
+	const onPayload: typeof threaded.onPayload = async (payload, model) => {
+		const prepared = await threaded.onPayload(payload, model);
+		const sanitized: Record<string, unknown> = {
+			...(prepared as Record<string, unknown>),
+			store: true,
+			parallel_tool_calls: false,
+		};
+		delete sanitized.previous_response_id;
+		if (threaded.previousResponseId) sanitized.previous_response_id = threaded.previousResponseId;
+		// CUA uses stored response state instead of stateless encrypted reasoning replay.
+		delete sanitized.include;
+		return sanitized;
+	};
+	return { context: threaded.context, onPayload };
+}
+
+// Meta implements the OpenAI Responses wire protocol. The model carries a
+// distinct api id so response ids can never be threaded across providers.
+export const streamMetaResponses: StreamFunction<typeof META_RESPONSES_API, MetaResponsesOptions> = (model, context, options) => {
+	const threaded = threadMetaRequest(context, options);
+	return piStreamOpenAIResponses(model as never, threaded.context, { ...options, onPayload: threaded.onPayload });
+};
+
+export const streamSimpleMetaResponses: StreamFunction<typeof META_RESPONSES_API, CuaSimpleStreamOptions> = (model, context, options) => {
+	const threaded = threadMetaRequest(context, options);
+	return piStreamSimpleOpenAIResponses(model as never, threaded.context, { ...options, onPayload: threaded.onPayload });
+};
diff --git a/packages/ai/src/providers/openai/provider.ts b/packages/ai/src/providers/openai/provider.ts
index b9fbfc86..36febae6 100644
--- a/packages/ai/src/providers/openai/provider.ts
+++ b/packages/ai/src/providers/openai/provider.ts
@@ -2,7 +2,6 @@ import {
 	type Context,
 	type OpenAIResponsesOptions as PiOpenAIResponsesOptions,
 	type StreamFunction,
-	type StreamOptions,
 } from "@earendil-works/pi-ai";
 import {
 	stream as piStreamOpenAIResponses,
@@ -10,41 +9,21 @@ import {
 } from "@earendil-works/pi-ai/api/openai-responses";
 import {
 	type CuaSimpleStreamOptions,
-	responseThreadingDelta,
-	responseThreadingEnabled,
-	type ResponseThreadingOptions,
+	type ResponsesThreadingOptions,
+	threadResponsesRequest,
 } from "../common";
 
 export const OPENAI_CUA_RESPONSES_API = "openai-cua-responses";
 
 /** Stream options for the cua OpenAI Responses provider: pi-ai's options plus threading control. */
-export interface OpenAIResponsesOptions extends PiOpenAIResponsesOptions, ResponseThreadingOptions {}
-
-type OnPayload = NonNullable<StreamOptions["onPayload"]>;
+export interface OpenAIResponsesOptions extends PiOpenAIResponsesOptions, ResponsesThreadingOptions {}
 
 /**
- * Prepare a request for pi-ai's builtin OpenAI Responses stream so it threads
- * `previous_response_id`. The public Responses API requires `store: true` to
- * chain, so the payload always stores; when a prior assistant `responseId`
- * exists, only the delta messages are sent with `previous_response_id` set.
- * Any caller `onPayload` runs on top of the threaded payload.
+ * Prepare a request for the OpenAI Responses transport with stored response state.
  */
-export function threadRequest(
-	context: Context,
-	options: (ResponseThreadingOptions & { onPayload?: OnPayload }) | undefined,
-): { context: Context; onPayload: OnPayload } {
-	const delta = responseThreadingEnabled(options) ? responseThreadingDelta(context.messages, OPENAI_CUA_RESPONSES_API) : undefined;
-	const previousResponseId = delta?.previousResponseId;
-	const messages = previousResponseId && delta ? delta.deltaMessages : context.messages;
-	const onPayload: OnPayload = async (payload, model) => {
-		const threaded = {
-			...(payload as Record<string, unknown>),
-			store: true,
-			...(previousResponseId ? { previous_response_id: previousResponseId } : {}),
-		};
-		return options?.onPayload ? ((await options.onPayload(threaded, model)) ?? threaded) : threaded;
-	};
-	return { context: messages === context.messages ? context : { ...context, messages }, onPayload };
+export function threadRequest(context: Context, options: ResponsesThreadingOptions | undefined) {
+	const { context: threadedContext, onPayload } = threadResponsesRequest(context, OPENAI_CUA_RESPONSES_API, options);
+	return { context: threadedContext, onPayload };
 }
 
 // pi-ai's builtin stream fns are typed to the "openai-responses" api; we reuse them under our routed api, hence `as never` on the model.
diff --git a/packages/ai/src/runtime-spec.ts b/packages/ai/src/runtime-spec.ts
index e8fbee66..867d4df7 100644
--- a/packages/ai/src/runtime-spec.ts
+++ b/packages/ai/src/runtime-spec.ts
@@ -5,6 +5,7 @@ import { modeForNativeTool, resolveNativeTool, type CuaNativeToolSpec } from "./
 import { providerModule as anthropic } from "./providers/anthropic/index";
 import { createNativeToolOnPayload, nativeApiForToolType, nativeToolExecutors } from "./providers/anthropic/native";
 import { providerModule as gemini } from "./providers/gemini/index";
+import { providerModule as meta } from "./providers/meta/index";
 import { providerModule as openai } from "./providers/openai/index";
 import { providerModule as tzafon } from "./providers/tzafon/index";
 import { providerModule as yutori } from "./providers/yutori/index";
@@ -20,6 +21,7 @@ const PROVIDERS = {
 	openai,
 	anthropic,
 	google: gemini,
+	meta,
 	tzafon,
 	yutori,
 } satisfies Record<CuaProvider, CuaProviderModule>;
diff --git a/packages/ai/test/api-keys.test.ts b/packages/ai/test/api-keys.test.ts
index 501b0106..3ab7c86f 100644
--- a/packages/ai/test/api-keys.test.ts
+++ b/packages/ai/test/api-keys.test.ts
@@ -12,6 +12,7 @@ const ENV_KEYS = [
 	"ANTHROPIC_API_KEY",
 	"GOOGLE_API_KEY",
 	"GEMINI_API_KEY",
+	"META_API_KEY",
 	"TZAFON_API_KEY",
 	"YUTORI_API_KEY",
 ] as const;
@@ -31,10 +32,12 @@ describe("cua api key helpers", () => {
 		expect(cuaApiKeyEnvVarsForProvider("openai")).toEqual(["OPENAI_API_KEY"]);
 		expect(cuaApiKeyEnvVarsForProvider("google")).toEqual(["GOOGLE_API_KEY", "GEMINI_API_KEY"]);
 		expect(cuaApiKeyEnvVarsForProvider("gemini")).toEqual(["GOOGLE_API_KEY", "GEMINI_API_KEY"]);
+		expect(cuaApiKeyEnvVarsForProvider("meta")).toEqual(["META_API_KEY"]);
 		expect(cuaApiKeyEnvVarsForProvider("unknown")).toEqual([]);
 	});
 
 	it("resolves provider api keys with fallback order", () => {
+		delete process.env.GOOGLE_API_KEY;
 		process.env.GEMINI_API_KEY = "gemini";
 		expect(getCuaEnvApiKey("google")).toBe("gemini");
 		process.env.GOOGLE_API_KEY = "google";
@@ -44,6 +47,8 @@ describe("cua api key helpers", () => {
 	it("resolves keys from model refs", () => {
 		process.env.OPENAI_API_KEY = "openai";
 		expect(getCuaEnvApiKeyForModel("openai:gpt-5.5")).toBe("openai");
+		process.env.META_API_KEY = "meta";
+		expect(getCuaEnvApiKeyForModel("meta:muse-spark-1.1")).toBe("meta");
 	});
 
 	it("throws readable errors when missing", () => {
diff --git a/packages/ai/test/computer-tool.integration.test.ts b/packages/ai/test/computer-tool.integration.test.ts
index 3aa06717..dcb59097 100644
--- a/packages/ai/test/computer-tool.integration.test.ts
+++ b/packages/ai/test/computer-tool.integration.test.ts
@@ -1,6 +1,6 @@
+import { existsSync } from "node:fs";
 import { readFile } from "node:fs/promises";
-import { dirname, join } from "node:path";
-import { fileURLToPath } from "node:url";
+import { join } from "node:path";
 import { describe, expect, it } from "vitest";
 import {
 	CUA_ACTION_TYPES,
@@ -10,14 +10,20 @@ import {
 	anthropic,
 	cuaModels,
 	gemini,
+	getCuaEnvApiKey,
 	getCuaModel,
+	meta,
 	openai,
 	tzafon,
 	yutori,
 } from "../src/index";
 
-const here = dirname(fileURLToPath(import.meta.url));
-const screenshotPath = join(here, "..", "examples", "screenshot.png");
+const screenshotPath = [
+	join(process.cwd(), "examples", "screenshot.png"),
+	join(process.cwd(), "packages", "ai", "examples", "screenshot.png"),
+].find(existsSync);
+
+if (!screenshotPath) throw new Error("could not find packages/ai/examples/screenshot.png");
 
 interface ProviderCase {
 	provider: CuaProvider;
@@ -105,9 +111,13 @@ async function buildYutoriContext(): Promise<Context> {
 	};
 }
 
+function apiKeyForCase(c: ProviderCase): string | undefined {
+	return c.provider === "meta" ? getCuaEnvApiKey("meta") : process.env[c.envVar];
+}
+
 describe("individual computer action integration", () => {
 	for (const c of cases) {
-		const hasKey = !!process.env[c.envVar];
+		const hasKey = !!apiKeyForCase(c);
 		const ciEnabled = !c.ciOptInEnvVar || !process.env.CI || process.env[c.ciOptInEnvVar] === "1";
 		const test = hasKey ? it : it.skip;
 
@@ -115,7 +125,7 @@ describe("individual computer action integration", () => {
 			const model = getCuaModel(c.modelRef as never);
 			const context = await buildContext(c.tools);
 			const response = await cuaModels().complete(model, context, {
-				apiKey: process.env[c.envVar],
+				apiKey: apiKeyForCase(c),
 				maxTokens: 1024,
 				...c.extraOptions,
 			});
@@ -142,6 +152,70 @@ describe("individual computer action integration", () => {
 		}, 60_000);
 	}
 
+	const metaApiKey = getCuaEnvApiKey("meta");
+	(metaApiKey ? it : it.skip)(
+		"meta continues a screenshot tool loop with previous_response_id",
+		async () => {
+			const screenshot = await readFile(screenshotPath);
+			const model = getCuaModel("meta:muse-spark-1.1");
+			const tools = meta.computerTools({ actions: ["click"] });
+			const context: Context = {
+				systemPrompt: "Use normalized 0-1000 coordinates. Call click exactly once, then answer in text after its result.",
+				messages: [
+					{
+						role: "user",
+						content: [
+							{ type: "text", text: "Click the sign in / up link. After the tool result, report that the click completed without calling another tool." },
+							{ type: "image", data: screenshot.toString("base64"), mimeType: "image/png" },
+						],
+						timestamp: Date.now(),
+					},
+				],
+				tools,
+			};
+			const first = await cuaModels().complete(model, context, {
+				apiKey: metaApiKey,
+				maxTokens: 1024,
+			});
+			const click = first.content.find((part) => part.type === "toolCall" && part.name === "click");
+			expect(click).toBeDefined();
+			expect(typeof click!.arguments.x).toBe("number");
+			expect(typeof click!.arguments.y).toBe("number");
+			expect(click!.arguments.x as number).toBeGreaterThanOrEqual(0);
+			expect(click!.arguments.x as number).toBeLessThanOrEqual(1000);
+			expect(click!.arguments.y as number).toBeGreaterThanOrEqual(0);
+			expect(click!.arguments.y as number).toBeLessThanOrEqual(1000);
+			expect(first.responseId).toBeTruthy();
+			context.messages.push(first, {
+				role: "toolResult",
+				toolCallId: click!.id,
+				toolName: click!.name,
+				content: [
+					{ type: "text", text: "Click executed successfully." },
+					{ type: "image", data: screenshot.toString("base64"), mimeType: "image/png" },
+				],
+				isError: false,
+				timestamp: Date.now(),
+			});
+
+			let payload: Record<string, unknown> | undefined;
+			const second = await cuaModels().complete(model, context, {
+				apiKey: metaApiKey,
+				maxTokens: 1024,
+				onPayload: (value) => {
+					payload = value as Record<string, unknown>;
+				},
+			});
+			expect(second.stopReason).not.toBe("error");
+			expect(second.responseId).toBeTruthy();
+			expect(payload?.previous_response_id).toBe(first.responseId);
+			expect(payload?.store).toBe(true);
+			expect(payload?.parallel_tool_calls).toBe(false);
+			expect(payload?.include).toBeUndefined();
+		},
+		90_000,
+	);
+
 	const yutoriHasKey = !!process.env.YUTORI_API_KEY;
 	(yutoriHasKey ? it : it.skip)(
 		"yutori translates native tool calls into canonical individual actions",
diff --git a/packages/ai/test/meta-threading.test.ts b/packages/ai/test/meta-threading.test.ts
new file mode 100644
index 00000000..43f18a16
--- /dev/null
+++ b/packages/ai/test/meta-threading.test.ts
@@ -0,0 +1,97 @@
+import { describe, expect, it } from "vitest";
+import type { Context, Message, Model } from "@earendil-works/pi-ai";
+import { META_RESPONSES_API, threadMetaRequest } from "../src/providers/meta/provider";
+
+const model = {} as Model<typeof META_RESPONSES_API>;
+
+function multiTurnContext(): Context {
+	const messages: Message[] = [
+		{ role: "user", content: "inspect the browser", timestamp: 0 },
+		{
+			role: "assistant",
+			content: [{ type: "toolCall", id: "call_1|fc_1", name: "screenshot", arguments: {} }],
+			api: META_RESPONSES_API,
+			provider: "meta",
+			model: "muse-spark-1.1",
+			responseId: "resp_meta_1",
+			usage: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, totalTokens: 0, cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 } },
+			stopReason: "toolUse",
+			timestamp: 0,
+		},
+		{
+			role: "toolResult",
+			toolCallId: "call_1|fc_1",
+			toolName: "screenshot",
+			content: [{ type: "image", mimeType: "image/png", data: "screenshot" }],
+			isError: false,
+			timestamp: 0,
+		},
+	];
+	return { messages, tools: [], systemPrompt: "control the browser" };
+}
+
+describe("Meta Responses threading", () => {
+	it("threads the latest response and applies serial computer-use defaults", async () => {
+		const { context, onPayload } = threadMetaRequest(multiTurnContext(), undefined);
+		expect(context.messages).toHaveLength(1);
+		expect(await onPayload({ input: [] }, model)).toEqual({
+			input: [],
+			store: true,
+			parallel_tool_calls: false,
+			previous_response_id: "resp_meta_1",
+		});
+	});
+
+	it("omits encrypted reasoning replay on the initial request", async () => {
+		const context: Context = { messages: [{ role: "user", content: "hello", timestamp: 0 }] };
+		const { onPayload } = threadMetaRequest(context, undefined);
+		expect(await onPayload({ input: [], include: ["reasoning.encrypted_content"] }, model)).toEqual({
+			input: [],
+			store: true,
+			parallel_tool_calls: false,
+		});
+	});
+
+	it("sends full history without previous_response_id when threading is disabled", async () => {
+		const context = multiTurnContext();
+		const prepared = threadMetaRequest(context, {
+			disableResponseThreading: true,
+			onPayload: (payload) => ({ ...(payload as object), previous_response_id: "foreign-response" }),
+		});
+		expect(prepared.context).toBe(context);
+		expect(await prepared.onPayload({ input: [] }, model)).toEqual({
+			input: [],
+			store: true,
+			parallel_tool_calls: false,
+		});
+	});
+
+	it("passes the prepared request to a caller payload hook", async () => {
+		const { onPayload } = threadMetaRequest(multiTurnContext(), {
+			onPayload: (payload) => ({ ...(payload as object), caller_field: true }),
+		});
+		expect(await onPayload({}, model)).toEqual({
+			caller_field: true,
+			store: true,
+			parallel_tool_calls: false,
+			previous_response_id: "resp_meta_1",
+		});
+	});
+
+	it("reapplies Meta constraints after caller payload hooks", async () => {
+		const { onPayload } = threadMetaRequest(multiTurnContext(), {
+			onPayload: (payload) => ({
+				...(payload as object),
+				store: false,
+				parallel_tool_calls: true,
+				previous_response_id: "wrong-response",
+				include: ["reasoning.encrypted_content"],
+			}),
+		});
+		expect(await onPayload({}, model)).toEqual({
+			store: true,
+			parallel_tool_calls: false,
+			previous_response_id: "resp_meta_1",
+		});
+	});
+});
diff --git a/packages/ai/test/models.test.ts b/packages/ai/test/models.test.ts
index f383547a..622d73df 100644
--- a/packages/ai/test/models.test.ts
+++ b/packages/ai/test/models.test.ts
@@ -6,6 +6,7 @@ import {
 	formatCuaModelRef,
 	getCuaModel,
 	listCuaModels,
+	meta,
 	openai,
 	parseCuaModelRef,
 } from "../src/index";
@@ -24,7 +25,7 @@ describe("CUA model refs", () => {
 
 	it("names the valid providers in the unsupported-provider error", () => {
 		expect(() => parseCuaModelRef("bogus:model")).toThrow(
-			'unsupported CUA provider "bogus" (expected one of: openai, anthropic, google, tzafon, yutori)',
+			'unsupported CUA provider "bogus" (expected one of: openai, anthropic, google, meta, tzafon, yutori)',
 		);
 	});
 
@@ -50,6 +51,14 @@ describe("CUA model refs", () => {
 		const model = getCuaModel("yutori:n1.5-latest");
 		expect(model.provider).toBe("yutori");
 		expect(model.api).toBe("yutori-chat-completions");
+
+		const muse = getCuaModel("meta:muse-spark-1.1");
+		expect(muse.provider).toBe("meta");
+		expect(muse.api).toBe(meta.META_RESPONSES_API);
+		expect(muse.baseUrl).toBe("https://api.meta.ai/v1");
+		expect(muse.contextWindow).toBe(1_048_576);
+		expect(muse.maxTokens).toBe(128_000);
+		expect(muse.thinkingLevelMap?.off).toBeNull();
 	});
 
 	it("loads supported custom provider models without explicit registration", () => {
@@ -57,11 +66,12 @@ describe("CUA model refs", () => {
 		expect(getCuaModel("yutori:n1.5-latest").api).toBe("yutori-chat-completions");
 	});
 
-	it("routes OpenAI CUA models to the threading-aware openai-cua-responses api", () => {
+	it("routes Responses models to provider-specific threading APIs", () => {
 		// Registry models carry pi-ai's builtin "openai-responses" api and must
-		// be routed to cua's own previous_response_id-threading provider.
+		// be routed to provider-specific previous_response_id transports.
 		expect(getCuaModel("openai:gpt-5.5").api).toBe(openai.OPENAI_CUA_RESPONSES_API);
 		expect(getCuaModel("openai:gpt-5.4-mini").api).toBe(openai.OPENAI_CUA_RESPONSES_API);
+		expect(getCuaModel("meta:muse-spark-1.1").api).toBe(meta.META_RESPONSES_API);
 	});
 
 	it("rejects supported model IDs that are not in pi-ai or overrides", () => {
@@ -115,6 +125,7 @@ describe("CUA support annotations", () => {
 
 	it("matches exact-id annotations", () => {
 		expect(findCuaAnnotation("google", "gemini-3-flash-preview")).toBeDefined();
+		expect(findCuaAnnotation("meta", "muse-spark-1.1")).toBeDefined();
 		expect(findCuaAnnotation("google", "gemini-3.1-flash-lite")).toBeDefined();
 		expect(findCuaAnnotation("yutori", "n1.5-latest")).toBeDefined();
 		expect(findCuaAnnotation("tzafon", "tzafon.northstar-cua-fast")).toBeDefined();
diff --git a/packages/ai/test/provider-module.test.ts b/packages/ai/test/provider-module.test.ts
index 88e81af8..b3f48106 100644
--- a/packages/ai/test/provider-module.test.ts
+++ b/packages/ai/test/provider-module.test.ts
@@ -1,11 +1,12 @@
 import { describe, expect, it } from "vitest";
-import { anthropic, CUA_COMPUTER_ACTION_TYPES, CUA_PROVIDERS, type CuaProvider, gemini, openai, tzafon, yutori } from "../src/index";
+import { anthropic, CUA_COMPUTER_ACTION_TYPES, CUA_PROVIDERS, type CuaProvider, gemini, meta, openai, tzafon, yutori } from "../src/index";
 import type { CuaProviderModule } from "../src/providers/common";
 
 const MODULES: Record<CuaProvider, { providerModule: CuaProviderModule }> = {
 	openai,
 	anthropic,
 	google: gemini,
+	meta,
 	tzafon,
 	yutori,
 };
@@ -14,6 +15,7 @@ const NAMESPACES: Record<CuaProvider, { namespace: Record<string, unknown>; pref
 	openai: { namespace: openai, prefix: "OPENAI" },
 	anthropic: { namespace: anthropic, prefix: "ANTHROPIC" },
 	google: { namespace: gemini, prefix: "GEMINI" },
+	meta: { namespace: meta, prefix: "META" },
 	tzafon: { namespace: tzafon, prefix: "TZAFON" },
 	yutori: { namespace: yutori, prefix: "YUTORI" },
 };
@@ -51,6 +53,11 @@ describe("provider modules satisfy the uniform contract", () => {
 	it("yutori sends no model-facing tool definitions", () => {
 		expect(yutori.providerModule.toolDefinitions()).toEqual([]);
 	});
+
+	it("meta currently restricts its normalized coordinate contract to computer mode", () => {
+		expect(() => meta.providerModule.toolDefinitions({ mode: "browser" })).toThrow(/computer only/);
+		expect(() => meta.providerModule.toolDefinitions({ mode: "hybrid" })).toThrow(/computer only/);
+	});
 });
 
 describe("provider namespaces export a uniform surface", () => {
diff --git a/packages/ai/test/providers.test.ts b/packages/ai/test/providers.test.ts
index c2b89737..8eb90faa 100644
--- a/packages/ai/test/providers.test.ts
+++ b/packages/ai/test/providers.test.ts
@@ -8,7 +8,7 @@ const YUTORI_CHAT_COMPLETIONS_API = yutori.YUTORI_CHAT_COMPLETIONS_API;
 describe("createCuaModels", () => {
 	it("registers the CUA-only providers alongside pi's builtins", () => {
 		const models = createCuaModels();
-		for (const id of ["openai", "anthropic", "google", "tzafon", "yutori"]) {
+		for (const id of ["openai", "anthropic", "google", "meta", "tzafon", "yutori"]) {
 			const provider = models.getProvider(id);
 			expect(provider, id).toBeDefined();
 			expect(provider?.stream).toBeTypeOf("function");
@@ -16,8 +16,9 @@ describe("createCuaModels", () => {
 		}
 	});
 
-	it("lists the tzafon and yutori catalogs on their providers", () => {
+	it("lists the CUA-only provider catalogs", () => {
 		const models = createCuaModels();
+		expect(models.getModel("meta", "muse-spark-1.1")?.api).toBe("meta-responses");
 		const tzafonIds = models.getModels("tzafon").map((m) => m.id);
 		expect(tzafonIds).toContain("tzafon.northstar-cua-fast");
 		const yutoriIds = models.getModels("yutori").map((m) => m.id);
diff --git a/packages/ai/test/runtime-spec.test.ts b/packages/ai/test/runtime-spec.test.ts
index bd7ee938..4233339a 100644
--- a/packages/ai/test/runtime-spec.test.ts
+++ b/packages/ai/test/runtime-spec.test.ts
@@ -1,5 +1,5 @@
 import { describe, expect, it } from "vitest";
-import { CUA_NAVIGATION_TOOL_NAME, CUA_PROVIDERS, getCuaModel, listCuaModels, openai, resolveCuaRuntimeSpec } from "../src/index";
+import { CUA_NAVIGATION_TOOL_NAME, CUA_PROVIDERS, getCuaModel, listCuaModels, meta, openai, resolveCuaRuntimeSpec } from "../src/index";
 
 describe("resolveCuaRuntimeSpec", () => {
 	it("resolves a runtime spec for every CUA provider", () => {
@@ -45,9 +45,12 @@ describe("resolveCuaRuntimeSpec", () => {
 		expect(anthropicSpec.onPayload).toBeTypeOf("function");
 	});
 
-	it("routes a concrete OpenAI Model input (not just a string ref) to openai-cua-responses", () => {
-		const raw = { ...getCuaModel("openai:gpt-5.5"), api: "openai-responses" } as Parameters<typeof resolveCuaRuntimeSpec>[0];
-		expect(resolveCuaRuntimeSpec(raw).model.api).toBe(openai.OPENAI_CUA_RESPONSES_API);
+	it("routes concrete Responses models to provider-specific APIs", () => {
+		const openaiModel = { ...getCuaModel("openai:gpt-5.5"), api: "openai-responses" } as Parameters<typeof resolveCuaRuntimeSpec>[0];
+		expect(resolveCuaRuntimeSpec(openaiModel).model.api).toBe(openai.OPENAI_CUA_RESPONSES_API);
+
+		const metaModel = { ...getCuaModel("meta:muse-spark-1.1"), api: "openai-responses" } as Parameters<typeof resolveCuaRuntimeSpec>[0];
+		expect(resolveCuaRuntimeSpec(metaModel).model.api).toBe(meta.META_RESPONSES_API);
 	});
 
 	it("threads tool options through to the provider module", () => {
diff --git a/packages/cli/README.md b/packages/cli/README.md
index 2e122292..7debfac9 100644
--- a/packages/cli/README.md
+++ b/packages/cli/README.md
@@ -45,6 +45,7 @@ cua models -p openai
 cua --print --model openai:gpt-5.5 "..."
 cua --print --model anthropic:claude-opus-4-7 "..."
 cua --print --model google:gemini-3-flash-preview "..."
+cua --print --model meta:muse-spark-1.1 "..."
 cua --print --model yutori:n1.5-latest "..."
 
 # Named sessions (browser stays alive across calls):
@@ -67,8 +68,8 @@ cua --session abc12345                        # by id prefix
 
 Run `cua models` to list every supported `-m` / `--model` value and the
 provider it routes to. Filter by provider with `cua models -p openai`,
-`cua models -p anthropic`, `cua models -p google` (alias: `gemini`), or
-`cua models -p yutori`.
+`cua models -p anthropic`, `cua models -p google` (alias: `gemini`),
+`cua models -p meta`, or `cua models -p yutori`.
 
 `-m` / `--model` accepts a provider-qualified `provider:model` ref (e.g.
 `openai:gpt-5.5`) or a bare model id when it matches exactly one catalog
@@ -85,12 +86,14 @@ Configuration is by environment variable. There is no config file.
 | `ANTHROPIC_API_KEY`  | Anthropic API key (required when `-m anthropic:…`) |
 | `GOOGLE_API_KEY`     | Google API key (required when `-m google:…`)   |
 | `GEMINI_API_KEY`     | alias of `GOOGLE_API_KEY`                      |
+| `META_API_KEY`       | Meta Model API key (required when `-m meta:…`)  |
 | `TZAFON_API_KEY`     | Tzafon API key (required when `-m tzafon:…`)   |
 | `YUTORI_API_KEY`     | Yutori API key (required when `-m yutori:…`)   |
 | `KERNEL_BASE_URL`    | override Kernel base URL                       |
 | `OPENAI_BASE_URL`    | override OpenAI base URL                       |
 | `ANTHROPIC_BASE_URL` | override Anthropic base URL                    |
 | `GOOGLE_BASE_URL`    | override Google base URL                       |
+| `META_BASE_URL`      | override Meta Model API base URL               |
 | `TZAFON_BASE_URL`    | override Tzafon base URL                       |
 | `YUTORI_BASE_URL`    | override Yutori base URL                       |
 | `XDG_DATA_HOME`      | sessions dir base (defaults to `~/.local/share`) |
diff --git a/packages/cli/src/cli-harness.ts b/packages/cli/src/cli-harness.ts
index 2de6d209..64d2d2eb 100644
--- a/packages/cli/src/cli-harness.ts
+++ b/packages/cli/src/cli-harness.ts
@@ -60,7 +60,7 @@ Usage:
   cua models --json
 
 Options:
-  -p, --provider <id>  Filter by provider: openai | anthropic | google | gemini | tzafon | yutori
+  -p, --provider <id>  Filter by provider: openai | anthropic | google | gemini | meta | tzafon | yutori
       --json           Output JSON
   -h, --help           Show this help
 `;
diff --git a/packages/cli/src/cli.ts b/packages/cli/src/cli.ts
index e474b565..5ed45148 100644
--- a/packages/cli/src/cli.ts
+++ b/packages/cli/src/cli.ts
@@ -53,6 +53,7 @@ Options:
                                    openai:    openai:gpt-5.5
                                    anthropic: anthropic:claude-opus-4-7
                                    google:    google:gemini-3-flash-preview
+                                   meta:      meta:muse-spark-1.1
                                    tzafon:    tzafon:tzafon.northstar-cua-fast
                                    yutori:    yutori:n1.5-latest
       --thinking <level>         Thinking level: off | minimal | low | medium | high | xhigh
@@ -101,12 +102,14 @@ Environment:
   ANTHROPIC_API_KEY     Anthropic API key (required when -m anthropic:…)
   GOOGLE_API_KEY        Google API key (required when -m google:…)
   GEMINI_API_KEY        Alias for GOOGLE_API_KEY
+  META_API_KEY          Meta Model API key (required when -m meta:…)
   TZAFON_API_KEY        Tzafon API key (required when -m tzafon:…)
   YUTORI_API_KEY        Yutori API key (required when -m yutori:…)
   KERNEL_BASE_URL       Override Kernel base URL
   OPENAI_BASE_URL       Override OpenAI base URL
   ANTHROPIC_BASE_URL    Override Anthropic base URL
   GOOGLE_BASE_URL       Override Google base URL
+  META_BASE_URL         Override Meta Model API base URL
   TZAFON_BASE_URL       Override Tzafon base URL
   YUTORI_BASE_URL       Override Yutori base URL
   XDG_DATA_HOME         Sessions are stored under \$XDG_DATA_HOME/cua/sessions
diff --git a/packages/cli/src/harness.ts b/packages/cli/src/harness.ts
index 2ffdf30f..ac1dbb6a 100644
--- a/packages/cli/src/harness.ts
+++ b/packages/cli/src/harness.ts
@@ -45,7 +45,9 @@ export interface BuildCuaHarnessOptions {
 	models?: Models;
 	/** Maximum tool-result images included from message history per provider request. */
 	toolResultImageReplayLimit?: CuaAgentHarnessOptions["toolResultImageReplayLimit"];
-	/** Chain OpenAI and Tzafon requests through provider-stored response state. Defaults to true. */
+	/**
+	 * Chain OpenAI, Meta, and Tzafon requests through provider-stored response state. Defaults to true.
+	 */
 	responseThreading?: CuaAgentHarnessOptions["responseThreading"];
 	/** Optional CUA-level retries around each provider request. Disabled by default. */
 	retry?: CuaAgentHarnessOptions["retry"];
diff --git a/packages/cli/test/cli-executor.test.ts b/packages/cli/test/cli-executor.test.ts
index 6cfb791c..1063b6a8 100644
--- a/packages/cli/test/cli-executor.test.ts
+++ b/packages/cli/test/cli-executor.test.ts
@@ -22,6 +22,7 @@ const PROVIDER_ENV_KEYS = [
 	"ANTHROPIC_API_KEY",
 	"GOOGLE_API_KEY",
 	"GEMINI_API_KEY",
+	"META_API_KEY",
 	"TZAFON_API_KEY",
 	"YUTORI_API_KEY",
 ];
diff --git a/packages/cli/test/harness-models.test.ts b/packages/cli/test/harness-models.test.ts
index 9354f364..d58e852e 100644
--- a/packages/cli/test/harness-models.test.ts
+++ b/packages/cli/test/harness-models.test.ts
@@ -9,16 +9,22 @@ describe("resolveCuaModelRef", () => {
 
 	it("passes provider-qualified refs through", () => {
 		expect(resolveCuaModelRef("openai:gpt-5.5")).toBe("openai:gpt-5.5");
+		expect(resolveCuaModelRef("meta:muse-spark-1.1")).toBe("meta:muse-spark-1.1");
 	});
 
 	it("accepts bare ids when they match exactly one catalog entry", () => {
 		expect(resolveCuaModelRef("gpt-5.5")).toBe("openai:gpt-5.5");
+		expect(resolveCuaModelRef("muse-spark-1.1")).toBe("meta:muse-spark-1.1");
 	});
 
 	it("throws on unknown bare ids", () => {
 		expect(() => resolveCuaModelRef("does-not-exist")).toThrow(/unknown model/);
 	});
 
+	it("filters the Meta catalog", () => {
+		expect(listSupportedModels("meta").map((model) => model.ref)).toEqual(["meta:muse-spark-1.1"]);
+	});
+
 	it("treats 'gemini' as an alias for google when filtering", () => {
 		const fromGemini = listSupportedModels("gemini");
 		const fromGoogle = listSupportedModels("google");