The memory your agents read from.
Knowledge, rules and prompts in one place, searchable and served by the same API as your model calls.
An AI knowledge base for agents, not for chat histories
Agents need context: product data, client rules, tone of voice, approvals. Today that context usually lives in long system prompts that every project copies differently. In the Brain Repository it lives once, with links between notes. The agent queries search and gets exactly the passages it needs.
Because the repository lives in the same account as the LLM gateway, both run on one access: fetch context, call the model, bill costs per project.
What the repository does today
Knowledge graph, not folders
Notes with links and backlinks. A workspace graph is available as nodes and edges. /graph, /backlinks
Full-text search for agents
One request returns matching notes and skills. The prompt stays short and the context stays current. /search?q=
Versioned prompts
Prompts and templates in the repository get a version on every change and can be restored to any earlier state. From Plus.
Git mirror
Pull and push to GitHub, GitLab or Gitea. Your knowledge also lives in your own repository.
Team roles
Per-workspace permissions for read, write or admin. The same repository for people and agents. From Pro.
Fetch context, then ask the model
Two calls, one account: search in the Brain Repository returns the context, the gateway sends it to the model with a budget limit.
All endpoints with examples in the API docs.
Context in one place
Free €0, Plus €8/month, Pro €15/month. All prices include VAT, cancel monthly. Operated in Germany.