Frequently Asked Questions (FAQ) about UBOS MCP Server

Q: What is the UBOS MCP Server?

A: The UBOS MCP (Model Context Protocol) Server is a tool available on the UBOS Asset Marketplace that connects AI agents to your GitHub corpora. It enables efficient information retrieval, local indexing, and seamless integration for smarter AI applications.

Q: How does the MCP Server work?

A: The MCP Server uses a hybrid semantic and keyword search to retrieve relevant information from your documentation database. It indexes your data locally, allowing AI agents to query, list, and retrieve documentation using FastMCP tools.

Q: What are the key features of the MCP Server?

A: Key features include efficient chunk-level retrieval, seamless agent integration, local-first design for low latency, automatic reindexing on file changes, a modular architecture, and extensibility for adding new tools or corpora.

Q: What file types are supported by the MCP Server?

A: The MCP Server supports a wide range of file types, including .py, .ipynb, .md, .txt, .yaml, and .yml.

Q: How do I add my own documentation to the MCP Server?

A: You can add your own documentation by placing it in the corpora/ directory. The server will automatically reindex the data and make it available to AI agents.

Q: What is FastMCP?

A: FastMCP is a protocol that standardizes how applications provide context to LLMs. The UBOS MCP Server is built on FastMCP, enabling it to provide a robust tools API for seamless agent integration.

Q: What is the purpose of the embed_documents.py script?

A: The embed_documents.py script loads, chunks, and embeds your documents, preparing them for efficient retrieval by AI agents.

Q: What is the purpose of the swarms_server.py script?

A: The swarms_server.py script starts the MCP server and provides the FastMCP tools API for interacting with the indexed data.

Q: How do I troubleshoot ‘No valid existing index found’ when starting the server?

A: The server will prompt you to embed and index documents. Accept the prompt or run python embed_documents.py manually first.

Q: How do I force a reindex of the documentation?

A: Run python swarms_server.py --reindex or use the swarm_docs.reindex tool.

Q: Can I disable Chroma compaction?

A: Yes, you can disable Chroma compaction if you prefer by setting the environment variable CHROMA_COMPACTION_SERVICE__COMPACTOR__DISABLED_COLLECTIONS to swarms_docs.

Q: How does the MCP Server handle file changes in the corpora/ directory?

A: The MCP Server watches the corpora/ directory for any file changes and automatically triggers a live reindex.

Q: What are the available FastMCP tools?

A: Available tools include swarm_docs.search for searching documentation chunks, swarm_docs.list_files for listing indexed files, swarm_docs.get_chunk for retrieving specific chunks, swarm_docs.reindex for forcing a reindex, and swarm_docs.healthcheck for checking server status.

Q: How can I integrate the MCP Server with the Windsurf Cascade IDE?

A: You can plug directly into Windsurf Cascade by configuring the command and arguments for the swarms_server.py script in the IDE settings.

Q: What Python version is required to run the MCP Server?

A: Python 3.11 is required.

Q: Where can I find more information about UBOS and its platform?

A: Visit the UBOS website at https://ubos.tech for more information about the UBOS platform and its features.