Linear Brain

Your AI project manager for Linear. No MCP required.

Linear Brain gives small teams (2-5 people, no dedicated PM) an AI-powered command centre for their Linear board. It reads your workspace freely, proposes changes through a human-approval queue, and generates daily PM-style insights — all through a dark-themed dashboard.

Why this exists

• Your Linear workspace has MCP disabled (common on enterprise plans) — AI tools that rely on MCP can't connect
• You're a small team without a PM and want AI to fill the gap — board hygiene, workload analysis, ticket cleanup
• You want AI to make changes to Linear but need human oversight on every single write operation
• You want a daily briefing on project health without manually crunching numbers

What it does

Dashboard

Real-time project overview: story points completed this week, cycle progress, per-person workload breakdown, blockers, and stale issues. One click to refresh from Linear.

Proposals (Approval Queue)

Every write operation goes through a proposal queue. Review what the AI wants to change, see the reasoning, then approve or reject. Batch approve/reject for speed.

Two AI-powered actions built in:

• Tidy Drafts — finds all tickets tagged DRAFT, rewrites titles/descriptions to your conventions, fixes labels, suggests estimates. One button.
• Audit Board — scans the entire board for convention violations, missing info, related tickets worth linking, and label/priority mismatches. Reads comments too, so it won't flag things already clarified in discussion.

Insights (AI PM Briefing)

Hit "Generate Insight" and Opus analyses your entire board — every ticket, every comment, every person's workload. It produces a structured daily briefing:

• Summary — the single most important thing to know today
• Cycle Progress — are you on track?
• Focus Areas — top 3 things that need attention right now
• Team Performance — per-person breakdown with candid assessments (who's overloaded, who's blocked, who's underutilised)
• Risks & Blockers — specific tickets that could derail the week
• Recommendations — concrete actions to take today

Designed to run every morning. Brutally honest, data-driven, repeatable. Previous insights are kept so you can track how things evolve.

Audit Log

Dev-only view of every action taken through the system.

How it works

Linear API ←── reader.ts (free reads)
                    ↓
         Dashboard snapshots / Board data
                    ↓
    ┌───────────────┼───────────────┐
    ↓               ↓               ↓
Tidy Drafts    Audit Board    Generate Insight
(headless CC)  (headless CC)  (headless CC)
    ↓               ↓               ↓
 Proposals       Proposals      Insight report
    ↓               ↓           (stored, displayed)
    └───────┬───────┘
            ↓
    Human reviews on dashboard
    Approve → Executor → writer.ts → Linear API
    Reject  → Logged

Headless actions use claude -p --model opus — the server gathers all board data (issues, labels, states, comments), sends it to Claude with a structured prompt, and processes the output. No Anthropic API SDK needed — just the Claude CLI.

Tech stack

Layer	Technology
Runtime	Bun
Language	TypeScript (strict)
Linear	@linear/sdk (official GraphQL SDK)
API server	Hono (JSON API)
Frontend	React + Ant Design (dark theme), built by Vite
Database	SQLite via bun:sqlite
AI	Claude Code CLI (headless Opus for actions/insights)

Setup

Prerequisites

• Bun (latest stable)
• Claude Code CLI installed and on PATH (for Tidy Drafts, Audit Board, and Insights)
• A Linear API key (Settings > API > Personal API keys)

Install

bun install

Configure

cp .env.example .env

LINEAR_API_KEY=lin_api_xxxxx
PORT=3000
DATABASE_PATH=./data/brain.db

Set up your board conventions:

cp BOARD_RULES.example.md BOARD_RULES.md
# Edit BOARD_RULES.md to match your team's naming, labels, and description format

Run

# Development (Vite HMR + API server with auto-reload)
bun run dev

# Production (builds frontend, then starts server)
bun run start

• Dev mode: Frontend on http://localhost:5173 (Vite, with HMR), API on http://localhost:3000
• Production: Everything on http://localhost:3000

Test

bun test

API

Method	Endpoint	Description
GET	/api/proposals	List proposals (filter by ?status=)
GET	/api/proposals/:id	Get a single proposal
POST	/api/proposals	Create a proposal
POST	/api/proposals/approve-all	Approve and execute all pending
POST	/api/proposals/reject-all	Reject all pending
POST	/api/proposals/:id/approve	Approve and execute a proposal
POST	/api/proposals/:id/reject	Reject a proposal
GET	/api/dashboard	Latest dashboard snapshot
POST	/api/dashboard/snapshot	Refresh dashboard data from Linear
GET	/api/insights	List all insights
POST	/api/insights/generate	Generate a new AI insight
POST	/api/actions/clean-drafts	Tidy all DRAFT tickets
POST	/api/actions/audit-board	Audit the board for issues
GET	/api/audit	Audit log entries
POST	/webhooks/linear	Linear webhook receiver

Safety model

• reader.ts can query Linear freely — no restrictions on reads
• writer.ts is the only module that calls Linear mutation methods
• executor.ts is the only module that imports writer.ts
• The executor verifies status === 'approved' before executing any proposal
• Headless AI actions create proposals — they never write to Linear directly
• All actions are logged to an audit table

Licence

MIT