Build and query a persistent, self-maintaining knowledge base that AI agents can reference and update locally. Ideal for developers who need agents to remember facts across conversations without external databases.
io.github.joshua-auchincloss/factsets
Local install
STDIO
No auth required
How models use it and what it is built for.
Build and query a persistent, self-maintaining knowledge base that AI agents can reference and update locally. Ideal for developers who need agents to remember facts across conversations without external databases.
Local install — runs as a subprocess.
Where to find authoritative docs and source for Factsets.
Paste any of these into Agent Studio after connecting Factsets.
Common questions about connecting and running Factsets.
What is a factset and how does it differ from a regular database?
A factset is a local, self-maintaining knowledge base designed for AI agents to store and retrieve facts across sessions. Unlike traditional databases, it's optimized for agent workflows and requires no external infrastructure—everything runs locally on your machine.
How do I install and get started with the factsets MCP server?
Install via npm with `npx factsets@0.2.0`. The server runs over stdio transport, so you can connect it to any MCP-compatible client. Check the project repository for configuration details and examples.
Is the knowledge base persistent across agent sessions?
Yes. Factsets maintains a persistent local store, so facts added by your agent in one session remain available in future sessions. The self-maintaining aspect means the knowledge base can clean up or organize itself automatically.
Can I use factsets in production or is it just for development?
Factsets is designed as a local knowledge base, making it ideal for development, testing, and single-machine deployments. For distributed or high-availability production systems, you may need to evaluate integration with external persistence layers.
What are the limitations of a local factset compared to cloud-based alternatives?
Local factsets don't require network calls or external services, making them faster and simpler to set up. However, they're limited to a single machine and don't offer built-in replication or multi-user access—cloud alternatives are better for team collaboration.
MCP Playground runs 10,000+ hosted MCP servers — GitHub, Linear, Notion, Stripe, Sentry and more — across Claude, GPT, Gemini, DeepSeek and 30+ AI models. Compare model answers side-by-side, save agent presets, share runs. Zero install.
Open Agent Studio