feat(sdk): add Strands Agents adapter

.github/workflows/ci.yml

@@ -7,6 +7,10 @@ on:
   # against the queued merge result. Without this the queue would wait forever.
   merge_group:
 
+# Every job here only reads the repo; nothing uses GITHUB_TOKEN for writes.
+permissions:
+  contents: read
+
 jobs:
 
   lint:

.github/workflows/e2e.yml

@@ -74,6 +74,11 @@ on:
 # the orphan sweep only reaps agents older than BAND_E2E_ORPHAN_MAX_AGE_MINUTES,
 # so a concurrent run's fresh agents are never swept out from under it.
 
+# Every job here only reads the repo. The GitHub API work uses E2E_GITHUB_TOKEN,
+# and the scorecard artifacts are downloaded from this same run.
+permissions:
+  contents: read
+
 jobs:
 
   lanes:

AGENTS.md

@@ -4,7 +4,7 @@ This is a Python SDK that connects AI agents to the Band collaborative platform.
 
 ## Core Features
 
-1. Multi-framework support (LangGraph, Anthropic, CrewAI, Claude SDK, Copilot SDK, Codex, Pydantic AI, Parlant, Gemini, Letta, Google ADK, OpenCode, Agno)
+1. Multi-framework support (LangGraph, Anthropic, CrewAI, Claude SDK, Copilot SDK, Codex, Pydantic AI, Parlant, Gemini, Letta, Google ADK, OpenCode, Agno, Strands Agents)
 2. A2A protocol support: Bridge to remote A2A agents and expose Band peers as A2A endpoints
 3. ACP integration: Editor-facing server and client adapters over stdio or TCP (Cursor, Codex, Claude Code, GitHub Copilot)
 4. Platform tools for chat, contacts, and memory management

README.md

@@ -330,6 +330,7 @@ For the full picture, rooms, contacts, platform tools, and how messages flow - s
 | Parlant          | `parlant`     | `ParlantAdapter`                     | | [examples](examples/parlant/)         |
 | Letta            | `letta`       | `LettaAdapter`                       | | [examples](examples/letta/)             |
 | Agno             | `agno`        | `AgnoAdapter`                        | | [examples](examples/agno/)              |
+| Strands Agents   | `strands`     | `StrandsAdapter`                     | | [examples](examples/strands/)         |
 | Codex            | `codex`       | `CodexAdapter`                       | [docs](docs/adapters/codex.md) | [examples](examples/codex/)             |
 | OpenCode         | `opencode`    | `OpencodeAdapter`                    | | [examples](examples/opencode/)       |
 
@@ -437,6 +438,7 @@ Adapter emit support:
 | Google ADK | Yes | - | - |
 | Pydantic AI | Yes | - | - |
 | LangGraph | Yes | - | - |
+| Strands Agents | Yes | - | - |
 | Parlant | - | - | - |
 | A2A / A2A Gateway | - | - | - |
 | ACP Client | - | - | - |
@@ -753,7 +755,7 @@ uv run python examples/run_agent.py --example anthropic
 uv run python examples/run_agent.py --example codex
 ```
 
-`examples/run_agent.py` supports `langgraph`, `pydantic_ai`, `anthropic`, `claude_sdk`, `parlant`, `crewai`, `codex`, `a2a`, and `a2a_gateway`, plus contact-management variants. Other supported adapters have direct example files: `examples/gemini/01_basic_agent.py`, `examples/google_adk/01_basic_agent.py`, `examples/letta/01_basic_agent.py`, `examples/agno/01_basic_agent.py`, and `examples/opencode/01_basic_agent.py`.
+`examples/run_agent.py` supports `langgraph`, `pydantic_ai`, `anthropic`, `claude_sdk`, `parlant`, `crewai`, `codex`, `a2a`, and `a2a_gateway`, plus contact-management variants. Other supported adapters have direct example files: `examples/gemini/01_basic_agent.py`, `examples/google_adk/01_basic_agent.py`, `examples/letta/01_basic_agent.py`, `examples/agno/01_basic_agent.py`, `examples/strands/01_basic_agent.py`, and `examples/opencode/01_basic_agent.py`.
 
 For a multi-framework collaboration demo that puts CrewAI agents and A2A-bridged services in the same room, see [examples/mixed](examples/mixed/).
 

agent_config.yaml.example

@@ -198,6 +198,19 @@ google_adk_agent:
   agent_id: ""
   api_key: ""
 
+# =============================================================================
+# Strands Agents Examples
+# =============================================================================
+
+# 01_basic_agent.py, 02_custom_tools.py, 03_custom_instructions.py,
+# 04_native_tools.py, 05_bedrock_model.py
+strands_agent:
+  agent_id: ""
+  api_key: ""
+
+# 06_tom_agent.py - see tom_agent in Shared Agents section
+# 07_jerry_agent.py - see jerry_agent in Shared Agents section
+
 # =============================================================================
 # Letta Examples
 # =============================================================================
@@ -304,13 +317,13 @@ arena_guesser_3:
 # =============================================================================
 
 # Tom the cat character agent
-# Used by: langgraph/07, claude_sdk/03, anthropic/03, pydantic_ai/03, parlant/04, crewai/05, agno/03, codex/02, opencode/05
+# Used by: langgraph/07, claude_sdk/03, anthropic/03, pydantic_ai/03, parlant/04, crewai/05, agno/03, codex/02, opencode/05, strands/06
 tom_agent:
   agent_id: ""
   api_key: ""
 
 # Jerry the mouse character agent
-# Used by: langgraph/08, claude_sdk/04, anthropic/04, pydantic_ai/04, parlant/05, crewai/06, agno/04, codex/03, opencode/06
+# Used by: langgraph/08, claude_sdk/04, anthropic/04, pydantic_ai/04, parlant/05, crewai/06, agno/04, codex/03, opencode/06, strands/07
 jerry_agent:
   agent_id: ""
   api_key: ""

examples/strands/01_basic_agent.py

@@ -0,0 +1,69 @@
+# /// script
+# requires-python = ">=3.11"
+# dependencies = ["band-sdk[strands]"]
+#
+# [tool.uv.sources]
+# band-sdk = { git = "https://github.com/band-ai/band-sdk-python.git" }
+# ///
+"""
+Basic Strands Agents example.
+
+This is the simplest way to create a Band agent with AWS Strands Agents.
+The adapter handles tool registration automatically.
+
+Strands has no provider-prefix model string (a bare string means a Bedrock
+model id), so the OpenAI provider is constructed explicitly; it reads
+OPENAI_API_KEY from the environment.
+
+Run with:
+    uv run examples/strands/01_basic_agent.py
+"""
+
+from __future__ import annotations
+
+import asyncio
+import logging
+import os
+
+from dotenv import load_dotenv
+from strands.models.openai import OpenAIModel
+
+from setup_logging import setup_logging
+from band import Agent
+from band.adapters import StrandsAdapter
+
+setup_logging()
+logger = logging.getLogger(__name__)
+
+
+async def main() -> None:
+    load_dotenv()
+
+    ws_url = os.getenv("BAND_WS_URL")
+    rest_url = os.getenv("BAND_REST_URL")
+
+    if not ws_url:
+        raise ValueError("BAND_WS_URL environment variable is required")
+    if not rest_url:
+        raise ValueError("BAND_REST_URL environment variable is required")
+
+    # Create adapter with framework-specific settings
+    adapter = StrandsAdapter(
+        model=OpenAIModel(model_id="gpt-5.4-mini"),
+        custom_section="You are a helpful assistant. Be concise and friendly.",
+    )
+
+    # Create and start agent
+    agent = Agent.from_config(
+        "strands_agent",
+        adapter=adapter,
+        ws_url=ws_url,
+        rest_url=rest_url,
+    )
+
+    logger.info("Starting Strands agent...")
+    await agent.run()
+
+
+if __name__ == "__main__":
+    asyncio.run(main())

examples/strands/02_custom_tools.py

@@ -0,0 +1,81 @@
+# /// script
+# requires-python = ">=3.11"
+# dependencies = ["band-sdk[strands]"]
+#
+# [tool.uv.sources]
+# band-sdk = { git = "https://github.com/band-ai/band-sdk-python.git" }
+# ///
+"""
+Strands agent with custom tools and capabilities.
+
+Shows the portable ``CustomToolDef`` (InputModel, handler) form shared across
+adapters, plus enabling memory/contact tools and execution/usage event
+emission via AdapterFeatures.
+
+Run with:
+    uv run examples/strands/02_custom_tools.py
+"""
+
+from __future__ import annotations
+
+import asyncio
+import logging
+import os
+
+from dotenv import load_dotenv
+from pydantic import BaseModel
+from strands.models.openai import OpenAIModel
+
+from setup_logging import setup_logging
+from band import Agent
+from band.adapters import StrandsAdapter
+from band.core.types import AdapterFeatures, Capability, Emit
+
+setup_logging()
+logger = logging.getLogger(__name__)
+
+
+class WeatherInput(BaseModel):
+    """Get the weather for a city."""
+
+    city: str
+
+
+async def get_weather(args: WeatherInput) -> str:
+    return f"{args.city}: sunny, 22°C"
+
+
+async def main() -> None:
+    load_dotenv()
+
+    ws_url = os.getenv("BAND_WS_URL")
+    rest_url = os.getenv("BAND_REST_URL")
+
+    if not ws_url:
+        raise ValueError("BAND_WS_URL environment variable is required")
+    if not rest_url:
+        raise ValueError("BAND_REST_URL environment variable is required")
+
+    adapter = StrandsAdapter(
+        model=OpenAIModel(model_id="gpt-5.4-mini"),
+        custom_section="You can check the weather with the weather tool.",
+        additional_tools=[(WeatherInput, get_weather)],  # CustomToolDef tuple
+        features=AdapterFeatures(
+            emit=frozenset({Emit.EXECUTION, Emit.USAGE}),
+            capabilities=frozenset({Capability.MEMORY, Capability.CONTACTS}),
+        ),
+    )
+
+    agent = Agent.from_config(
+        "strands_agent",
+        adapter=adapter,
+        ws_url=ws_url,
+        rest_url=rest_url,
+    )
+
+    logger.info("Starting Strands agent with custom tools...")
+    await agent.run()
+
+
+if __name__ == "__main__":
+    asyncio.run(main())

examples/strands/03_custom_instructions.py

@@ -0,0 +1,88 @@
+# /// script
+# requires-python = ">=3.11"
+# dependencies = ["band-sdk[strands]"]
+#
+# [tool.uv.sources]
+# band-sdk = { git = "https://github.com/band-ai/band-sdk-python.git" }
+# ///
+"""
+Strands agent with a fully custom system prompt.
+
+Two levers shape the prompt:
+
+* ``custom_section`` (see 01_basic_agent.py) is appended to the prompt the SDK
+  renders, so the Band tool contract stays in place;
+* ``system_prompt`` REPLACES that rendered prompt entirely — nothing about the
+  platform tools is injected for you, so the prompt below states the messaging
+  contract itself. Without it the model answers in plain text, the reply never
+  reaches the room, and the adapter reports a dropped-reply error.
+
+Run with:
+    uv run examples/strands/03_custom_instructions.py
+"""
+
+from __future__ import annotations
+
+import asyncio
+import logging
+import os
+
+from dotenv import load_dotenv
+from strands.models.openai import OpenAIModel
+
+from setup_logging import setup_logging
+from band import Agent
+from band.adapters import StrandsAdapter
+from band.core.types import AdapterFeatures, Emit
+
+setup_logging()
+logger = logging.getLogger(__name__)
+
+SUPPORT_PROMPT = """
+You are a technical support agent for a software company, working inside a Band
+chat room.
+
+How to reply:
+- Every reply to the room MUST go through the band_send_message tool, mentioning
+  the person you are answering. Plain text answers never reach the room.
+- Send exactly one message per turn.
+
+Guidelines:
+- Ask for the environment (OS, version, exact error) before troubleshooting.
+- Give numbered, verifiable steps.
+- Escalate to a human when the issue needs account or billing access.
+"""
+
+
+async def main() -> None:
+    load_dotenv()
+
+    ws_url = os.getenv("BAND_WS_URL")
+    rest_url = os.getenv("BAND_REST_URL")
+
+    if not ws_url:
+        raise ValueError("BAND_WS_URL environment variable is required")
+    if not rest_url:
+        raise ValueError("BAND_REST_URL environment variable is required")
+
+    adapter = StrandsAdapter(
+        model=OpenAIModel(model_id="gpt-5.4-mini"),
+        # Full override: custom_section would be ignored alongside this.
+        system_prompt=SUPPORT_PROMPT,
+        # Post each tool call and result into the room for visibility.
+        features=AdapterFeatures(emit={Emit.EXECUTION}),
+    )
+
+    agent = Agent.from_config(
+        "strands_agent",
+        adapter=adapter,
+        ws_url=ws_url,
+        rest_url=rest_url,
+    )
+
+    logger.info("Starting Strands support agent...")
+    await agent.run()
+
+
+if __name__ == "__main__":
+    asyncio.run(main())