Sailbit for agents
Connect Sailbit to your AI agent and run your workflow — capture work, get your next move, and close the loop — without leaving your terminal.
What this is
Sailbit exposes a Model Context Protocol (MCP) server so AI agents like Claude Code, Claude Desktop, and Cursor can work with your tasks and products. The agent acts as an advisor inside Sailbit's workflow — it captures and updates work for you, but it doesn't manage your backlog or change your strategy.
The loop
Agent access is built around Sailbit's core loop:
- Capture — drop in a task and classify it as build, distribution, or other — without switching apps.
- Decide — ask what to work on next and get Sailbit's single recommendation with the reason.
- Do — focus on the work itself.
- Close — mark it done or snooze it, right from the agent.
Security
Agent access is designed to be safe by default:
- Each API key maps to one account. Agents can only ever read and write your own data.
- Keys are hashed at rest, shown once, and revocable anytime from Settings.
- Tools only read and write tasks and weekly intents — never billing, account, or product settings.
- Requests are rate-limited per key, and your task content is returned to the agent as data, never as instructions.
Agent access is part of Sailbit Pro. Create an API key in Settings → Agent access.