DocsGPT
Open-source AI agents grounded in your docs — private, self-hosted, any model.
About
DocsGPT builds AI agents that answer from your documentation. It's private and self-hostable, works with any model (local or cloud), and is aimed at teams who want a chat interface over their product docs, wikis and knowledge bases.
Agents get tools (web search, REST APIs, MCP servers, sandboxed code), can be embedded in your product via a widget, and are exposed through an OpenAI-compatible API, an Agent API and webhooks. MIT licensed and actively maintained.
What it can do
- RAG chat over your documentation
- Custom agents with tools and MCP servers
- OpenAI-compatible API and webhooks
- Embeddable chat widget
- Works with local or cloud models
Setup
- Install Docker.
- Run the DocsGPT installer (it checks for Docker and installs the
docsgptcommand), or use Docker Compose directly. - Open the app and connect your documentation sources.
- Chat with your docs, or embed the agent via the widget or API.
Prerequisites
Docker; a model endpoint (local via Ollama or a cloud API key).
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: 18.3k stars — respectable for docs-RAG niche; pushed 2026-10-12 — active. Manual review: OpenAI-compatible API plus Agent API verified; MCP server and sandbox tool support show serious agent ambitions. License: MIT.
★ 18,319 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.0/10
- Maintenance · recency, releases, issue load
- 9.0/10
- Openness · license + docs
- 9.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · 18.3k stars — respectable for docs-RAG niche
- · pushed 2026-10-12 — active
- · manual review: OpenAI-compatible API plus Agent API verified; MCP server and sandbox tool support show serious agent ambitions
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.