All member projects

Code · Built by members

Braintrust

Builds AI personas from what people publish on Substack, YouTube, blogs and Bluesky, and rebuilds them as new posts arrive. Any MCP client can then consult them.

Conceptual illustration: Builds AI personas from what people publish on Substack, YouTube, blogs and Bluesky, and rebuilds them as new posts arrive. Any MCP client can then consult them.
AI-generated editorial illustration, not a product screenshot.

What it is

Braintrust builds AI personas of the people you choose to follow, based on what they have actually published. A static prompt that says respond like someone goes stale; Braintrust keeps each persona current and cites its answers back to dated sources.

It is a personal tool that you clone and deploy yourself, not a hosted service. It is built on the Open Brain (OB1) project and runs against your own database and your own model endpoints.

What it does

  • Reads Substack archives, YouTube captions, any blog with a feed or sitemap, and Bluesky posts a day at a time.
  • Reads back twelve months, then checks daily for new items.
  • Reads each item once, keeps notes, and rebuilds a persona only when something new arrives.
  • Keeps only quotes it can find word for word in the source.
  • Serves personas over MCP to clients such as Claude, ChatGPT and Cursor, and can give each person their own Hermes agent.
  • Includes an eval harness to test which model should read your corpus.

Built with

  • TypeScript
  • Node.js
  • Open Brain (OB1)
  • Supabase
  • Postgres
  • pgvector
  • MCP
  • Python
  • Shell
  • agents

More from the community

Demo

One neuron at a time

by Caleb

Push single neurons in small open models and turn them into villains, flatterers or pirates. Try the recorded replies, then run it yourself.

Code

governed-agent

by Jeremy

Capability-based security, scope-bounded planning and out-of-band audit for autonomous LLM agents.

Website

Dave, a personal agent

by Chris

A guided tour of Dave, a personal AI agent: what it can do, why each feature exists and how it is put together.

Back to all member projects