MCP Documentation Server
A local document base for your agent: hybrid semantic and keyword search plus a web UI for browsing and uploads
- 343 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The server reads and writes files in its data folder and keeps a local web port open
- With a GEMINI_API_KEY the document text goes to Gemini
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/mcp-documentation-serverDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The server reads and writes files in its data folder and keeps a local web port open
- With a GEMINI_API_KEY the document text goes to Gemini
This is third-party code. Review the repository files before installing.
What it does
MCP Documentation Server keeps your documents on your computer and lets an agent search them. Documents are added with the add_document tool or dropped into the uploads folder as txt, md and pdf files, after which process_uploads splits them into chunks and computes vectors. The Orama vector database and the Transformers.js embedding model run locally, so no external database or cloud API is needed. Search is hybrid: full text plus vector similarity. The search_all_documents tool searches every document at once, search_documents searches one, and get_context_window returns neighboring chunks around a hit so the model gets a wider context. If you set GEMINI_API_KEY, search_documents_with_ai also appears for document analysis by Gemini. A web interface starts on port 3080 next to the MCP server: a document list, drag-and-drop upload, search and chunk browsing. Every tool is also available as a REST API, and the repository contains a documentation-server skill that teaches an agent to call those endpoints without loading MCP schemas into context.
Who it is for. For developers who want to give Claude Code, Cursor or Gemini CLI search over the docs of a new framework, an API or internal guides without cloud services.
Good fit when
- You are learning a new framework and the agent needs its docs, which are missing from the model's training data
- Internal guides and specs live in md and pdf files and cannot go to an external cloud service
- You need search that finds both the exact word and a semantically close passage
- You want to browse and extend the knowledge base yourself in a browser
- You want to save context: the agent calls the REST API and skips loading MCP tool schemas
Not a fit when
- Your documents are docx, html or spreadsheets: the server will not take them
- You need a shared base for a team on several machines: the data sits in a local folder
- You need search over live sites and repositories: the server only works with what you uploaded
- The web UI port is taken by another service and you do not want to change settings
Example request
Process the files in the uploads folder, find the authentication section in them and show the neighboring chunksMCP
- Transport
- stdio
- Authentication
- not required
Environment variables
GEMINI_API_KEYsecret- Gemini key for the search_documents_with_ai tool
MCP_BASE_DIR- Folder for stored data
MCP_EMBEDDING_MODEL- Embedding model name
START_WEB_UI- Whether to start the web UI
WEB_PORT- Web UI port, 3080 by default
Limitations
By default data lives in the `~/.mcp-documentation-server` folder, and the MCP_BASE_DIR variable changes the path. All environment variables are optional.
The server is built for one computer with a local data folder: it has no shared store for a team. The default embedding model is Xenova/all-MiniLM-L6-v2, and the MCP_EMBEDDING_MODEL variable selects another. The skill talks to an already running server, so without Node.js and npx it is of no use. START_WEB_UI turns the web interface off, and WEB_PORT changes the port.
How to disable. Remove the documentation block from your MCP client configuration and restart the client. Uploaded documents stay in the data folder, which you can delete by hand.
FAQ
How do I install MCP Documentation Server in Claude Code?
Add a server to the MCP configuration that starts with `npx -y @andrea9293/mcp-documentation-server`. After launch the web interface opens at http://localhost:3080.
Do I need a Gemini key?
No, document search runs locally on the Xenova/all-MiniLM-L6-v2 model. The GEMINI_API_KEY variable is needed only for the search_documents_with_ai tool.
Which file formats does the server accept?
txt, md and pdf. Files go into the uploads folder, whose path get_uploads_path returns, and then you call process_uploads.
What does the documentation-server skill give me?
It describes every REST request of the server with examples. The agent calls them through curl, so MCP tool schemas take no space in the context.
How do I change the port or the data folder?
Through the WEB_PORT and MCP_BASE_DIR environment variables in the env block of the configuration. The web interface is turned off by setting START_WEB_UI to false.
Discussion
No questions or reviews yet.
Related
MCP serversAn MCP server with full-text search over local 1C:Element documentation, including your own markdown and PDF files
PageIndex MCP: long PDFs for an agent
PageIndex MCP
An MCP server that lets an agent read long PDFs through a document tree: retrieval by reasoning instead of a vector database
MCP PDF for large documents: the agent searches by meaning and keyword across one file or a folder and reads only the pages, tables and scans it needs
A local MCP server for arXiv literature work: section-level paper reads, BibTeX from metadata, topic watches
Semantic search over Jivo and SaluteBot docs
MCP Docs Hub
A collection of MCP servers with FAISS indexes for semantic search over Jivo, SaluteBot and SberSalute docs: search, file reads and contents
An MCP server built on Docling: converts PDFs into structured documents, builds new ones and feeds them to RAG
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| MCP Documentation Server (this entry) | no | not required | 343 | |
| MCP Local RAG | no | not required | 425 | local document search with no keys and no web UI |
| Docs MCP Server | no | not required | 1,840 | indexes the docs of websites, GitHub and packages |
| Context7 | yes | API key | 63,924 | a cloud service with up-to-date library docs |
README in short
The author stresses local operation with no external databases or cloud APIs and the built-in web interface. All twelve tools and the environment variables are laid out in tables.