AI-assisted generation
Generate new documentation from prompts, existing documents, and retrieved repository context instead of writing every page from scratch.
Free under MIT. Open source by default.
Documentation that compounds through AI.
Admont AI is a RAG-first documentation system. Teams create documentation with AI assistance, then use that documentation as the retrieval base for the next round of writing, answers, and knowledge work.
- MIT licensed
- AI-assisted documentation generation
- RAG on your own knowledge base
Product
A new documentation workflow, not just a new editor.
The core idea behind Admont AI is recursive documentation generation: AI helps create documentation, that documentation is stored as structured knowledge, and the accumulated knowledge base becomes the source for better future documentation. Git and Markdown are not the product story on their own. They are the best substrate for keeping that loop transparent, durable, and easy to evolve.
Features
Everything is built around the RAG authoring loop.
Compounding knowledge base
Each accepted document becomes material for later retrieval, so the system improves as the documentation corpus grows.
Hybrid retrieval
Combine full-text and semantic search to pull the most relevant context into answers, generation, and editing flows.
Markdown as working memory
Keep documentation in plain Markdown so generated knowledge stays durable, inspectable, and easy to reuse outside the app.
Git-backed review and history
Review, diff, publish, and recover generated content with a workflow that treats documentation as a real engineering asset.
MCP and assistant access
Expose the same retrieval-ready documentation base to external AI assistants through the Model Context Protocol.
How It Works
Generate, retrieve, refine, repeat.
Admont AI combines models with retrieval over your own repository so new documentation can be grounded in prior work instead of generated in a vacuum.
Generate
Writers and teams create or refine docs with integrated AI assistance.
Retrieve
Hybrid search brings the most relevant existing knowledge into the workflow.
Store
Approved output lives in Markdown and Git so it stays portable and auditable.
Reuse
The expanded corpus becomes the retrieval layer for the next generation cycle.
AI-Assisted Authoring
RAG Search
Markdown + Git
LLMs
MCP Reuse
Homage
Named in tribute to the Admont Abbey library.
Admont AI takes its name from the Admont Abbey library in Austria, a place shaped by the long human tradition of collecting, preserving, and sharing knowledge. The project is a modern homage to that idea: documentation as a living body of knowledge that can be expanded, retrieved, and passed forward.
Open Source
Free to use, inspect, and adapt.
Admont AI is MIT licensed. You can run it yourself, inspect the code, and adapt it to your own stack without a paid gate in the middle.