Bring your own agent

Most AI products bolt on a single assistant and hope it’s the one you want. Crilo takes the opposite bet: you already have an agent you trust, configured the way you like. Crilo just gives it a canvas to work on.

How it works

Every workspace exposes a remote MCP endpoint. Point any MCP-speaking client at it — Claude Code, Cursor, Windsurf, your own — and the agent can build on the board while you watch it happen live.

claude mcp add --transport http crilo https://mcp.crilo.app/mcp/<workspace-id>

That’s the whole setup. No plugin to install, no model to pick, no lock-in. Switch agents tomorrow and the canvas doesn’t care.

Why this matters

The best agent for your work is a moving target — it changes as models improve and as your own setup evolves. Tying your canvas to one built-in bot freezes you in time. Bringing your own keeps you on the frontier, automatically.