Allen is an agentic operating system for software development. These docs explain how to set up Allen, understand its main concepts, contribute safely, and operate the built-in agent/workflow system.
- First workflow - run Allen locally and launch a safe first workflow.
- Architecture - high-level system map and package responsibilities.
- Security and sandboxing - repo execution, credentials, public links, workspaces, and operator responsibilities.
- Troubleshooting - common setup, auth, MongoDB, Claude Code, workspace, integration, and test failures.
- Known limitations - current alpha constraints.
- Workflows - repeatable multi-step processes for agent work.
- Agents - configured AI workers with roles, tools, and permissions.
- Teams - the seeded agent organization and responsibility model.
- Skills - routing and operating guidance for choosing the right path.
- Workspaces - isolated repo worktrees for agent tasks.
- Artifacts - durable outputs from agents and workflows.
- Integrations - GitHub, Linear, Slack, MCP, and provider connections.
- Engine - workflow runtime and agent execution orchestration.
- Server - API, persistence, integrations, workspaces, and dispatch.
- UI - React app for chat, workflows, executions, workspaces, settings, and review.
- Desktop - Electron host for the shared Allen product.
- E2E - Playwright coverage for full-product flows.
- Context engine setup - install the optional context engine from the UI.
- Context engine installation - manual Python, model, Postgres/pgvector, and Neo4j setup details.
- Context engine - runtime context retrieval, mandatory injection, selection policy, and usage tracking.
- Curated and mandatory context - field-level guidance for creating and editing context entries.
- Self-healing monitoring - how Allen watches its own runtime and routes incidents.
- Desktop packaging - local macOS packaging and smoke validation.
- Documentation guidelines - how to keep public docs useful and high-level.
- Allen design system reference - visual language for Allen UI work.
- Context engine best practices - guidance for repo context quality.
- Claude prompting best practices - prompt guidance for Claude-backed agents.
- Gemini prompting best practices - prompt guidance for Gemini-backed agents.