Kwipu MCP
The MCP server answers questions from a personal document collection and returns linked passages with sources.
- 610 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The server reads selected documents and stores a graph; their contents reach the selected local or cloud model.
Install
- Works in
Manual install
{
"mcpServers": {
"kwipu": {
"command": "C:/path/to/Kwipu/.venv/Scripts/python.exe",
"args": ["C:/path/to/Kwipu/kwipu_mcp_server.py"],
"env": {
"KWIPU_KNOWLEDGE_DIR": "C:/path/to/vault",
"KWIPU_LLM_MODEL": "qwen2.5:7b",
"KWIPU_EMBED_MODEL": "nomic-embed-text"
}
}
}
}Configuration from the MCP server section. Install Kwipu dependencies and models first, then replace the example absolute paths.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The server reads selected documents and stores a graph; their contents reach the selected local or cloud model.
This is third-party code. Review the repository files before installing.
What it does
Kwipu MCP lets an agent query a document graph. Kwipu links notes through wikilinks, metadata and text meaning, then retrieves relevant passages for a question. The agent can request an answer alone or an answer with citations. The collection accepts Markdown, plain text, PDF and DOCX. Users can explore the same graph in a terminal or a separate web interface. Answers can use temporal metadata to trace decisions across meetings.
Who it is for. Project leads and documentation writers tracing decisions, roles and relationships in a note folder or Obsidian vault.
Good fit when
- Several meetings need to be traced back to the people and notes behind a decision.
- An Obsidian vault contains relationships between people and projects worth exploring.
- An agent answer needs source passages that can be opened and checked.
- Older meeting notes need to show how a decision changed over time.
Not a fit when
- You want the agent to edit source documents through this server.
- Source documents cannot be sent to the selected cloud model.
- The required information is outside the supported file formats.
Example request
Find Project Alpha scheduling decisions in Kwipu and show the notes supporting the answer.MCP
- Transport
- stdio
- Authentication
- not required
Limitations
Running the server requires Python 3.12 and Ollama with the selected models; check them before connecting an agent. The default language model may run in the cloud, so a local Ollama address alone does not keep documents on the machine. MCP builds missing storage but does not watch edits; keep the indexer running for updates. Changing the embedding model requires a rebuilt index because existing vectors belong to the previous model.
How to disable. Remove Kwipu from the MCP client and stop the indexer. Remove stored graph data only after all processes using it have stopped.
FAQ
How do I see what supports an answer?
Ask the agent to return citations with the answer. The server provides a detailed-query tool for this.
Why are new notes missing?
Check the configured document folder and running indexer. The MCP server itself does not watch file changes.
Can I work without a browser?
Yes. MCP and terminal use do not require the web interface.
What replaces the example paths?
Use absolute paths to your environment’s Python, the server script and document folder. Adapt the C-drive example to your installation.
What happens to notes with the same name?
Kwipu qualifies names with their folder paths so an active note can be distinguished from an archived namesake.
Discussion
No questions or reviews yet.
Related
MCP serversThe MCP server lets agents explore prepared documents, sections and related assets while retaining source references.
The MCP server lets agents read and search local Molio vaults, browse file trees and ask questions about their contents.
The MCP server connects an agent to a learning canvas where a question about a passage opens a linked document.
A local document base for your agent: hybrid semantic and keyword search plus a web UI for browsing and uploads
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
README in short
Kwipu records structural relation origins down to the file and line. Synonym retrieval is disabled in MCP mode.