Cursor

The repository tells you what was built. Never why.

What it changes

A coding agent reads your repository perfectly. It will never know that a feature was dropped after three customer calls, or that a technical choice came from a commercial constraint. That information was never in the code.

Connected to Verbasil, Cursor can read it before proposing anything.

What Verbasil brings to the conversation

Verbasil exposes about twenty read and write tools. In practice, the assistant can:

• retrieve what the company decided, learned, and what is still open;
• search the memory by topic, by person, by period;
• read a contact's context before drafting anything about them;
• propose a write: a settled decision, a learning, a report.

And the rule that makes all of this workable: a write is never applied directly. It arrives in Sources and waits for a human to review it. An assistant writing unchecked into a company's memory would make it useless within a month.

How to connect it

1. Open Verbasil, go to Integrations, AI tools section.
2. Generate an access token.
3. In Cursor, open the MCP settings.
4. Add a server with Verbasil's MCP address and your token.
5. Ask it why a given product decision was made before touching the code.

A read-only token is more than enough for this use.

What it does not do

Verbasil fetches nothing FROM Cursor: the direction is the opposite. Cursor queries the memory; the memory does not go through your conversations.

No write is applied without approval.

FAQ

Does Verbasil read my conversations?

No. Verbasil only ever sees the questions the assistant asks it, never the rest of the exchange.

Can I limit what the assistant is allowed to do?

Yes. A read-only token allows lookups and nothing else.

What happens if I revoke the token?

Access stops immediately. Whatever is already in the memory stays there.