- Enterprise Brain
- How it works
- What your agent sees
MCP tools
What your agent sees.
Seven MCP tools by default. Claude Code, Codex, Gemini CLI, Cursor and Copilot all connect the same way, and the agent decides when to call each one.
| Tool | What it does | When the agent calls it |
|---|---|---|
start_session | Opens the session. Returns what you and your team have saved: preferences and decisions to honour. | First, with what the user asked. |
get_advice | The note for this kind of task: the steps that worked, the pitfalls that sank earlier runs, and how many runs taught it. | Before substantive work; again when choosing an approach, when stuck, and when done. |
search_memory | Facts and passages, each with its source, filtered to what the reader may see. | Before guessing at what the team decided, said or prefers. |
get_episode | The source text a fact cites. | To read the meeting, thread or document behind a fact. |
remember_memory | Saves a preference, correction or decision. | When the user states one. |
forget_memory | Forgets something you saved, everywhere. | When the user takes it back. |
end_session | Closes the session with a short transcript and its outcome, so the advice it served is judged and a good outcome teaches it. | Last. |
Where a schema summarises a subject, such as a deal or an account, two more tools appear: get_brief returns that subject's brief and list_briefs lists them.