-
Notifications
You must be signed in to change notification settings - Fork 3
Home
John Williams edited this page Mar 9, 2026
·
2 revisions
Welcome to the ContextOS documentation. This wiki covers everything from quick-start installation to deep architecture internals.
| Section | What's Inside |
|---|---|
| Architecture | How the 5 layers fit together, data flow diagrams |
| Memory Layer | Tiering, entity graph, conflict resolution, persistence |
| Retrieval Layer | Hybrid search, multi-corpus routing, staleness detection |
| Tool Execution Layer | DAG pipelines, caching, retry policies, sandboxing |
| Planning & Spec Layer | Spec engine, sparring hook, dynamic revision, outcome eval |
| Orchestration Core | Intent router, schema registry, cost ledger, tracing |
| MCP Tools Reference | All 47 tools, schemas, parameters, examples |
| Credits & Origins | Full credit to all 6 source repos and what each contributed |
| Configuration | All config options, environment variables, workspace setup |
| Integrations | Claude Desktop, Cursor, Windsurf, LangChain, OpenAI SDK |
| Roadmap | Phase-by-phase build plan with status |
| Contributing | How to contribute, PR guidelines, code standards |
| FAQ | Common questions and sharp answers |
ContextOS is a unified MCP server that absorbs the capabilities of six leading open-source AI context projects — modelcontextprotocol/servers, ragflow, Prompt-Engineering-Guide, context7, claude-mem, composio, and get-shit-done — and builds the orchestration layer that none of them had. The result is 47 MCP tools across 5 layers: Memory, Retrieval, Tool Execution, Planning, and Orchestration. One pip install. pip install contextos.