Your AI tools, working from the project.

Anything that speaks MCP. In practice that is Claude, ChatGPT, Cursor and Codex. Each connects to the same projects, so it does not matter which one you open.

• Read a project. Its spaces, its assets and their statuses, and the facts the project treats as current.
• Search it by meaning, and open any single asset in full rather than working from an excerpt.
• Save what it writes (notes, scripts, prompts, briefs, images) into the project and the space you name.
• Create a project or a space when the work needs one, so a conversation can be filed without leaving the conversation.
• Propose what is true. An assistant can suggest a fact for the project’s state; it is marked as proposed and counts for nothing until you confirm it.

It cannot decide what the project treats as true. Proposals wait for you, and what you have approved outranks everything else it reads. It also cannot reach a project you have not connected it to, and it holds no access of its own: the connection is yours, made from Settings → Connections, and revoking it there ends it immediately.

Settings → Connections in Seylo creates an access token and shows the configuration for your tool. The step-by-step version, with what to do if the tool cannot see your projects afterwards, is in the help centre: connect Claude or ChatGPT through MCP.

A token acts as you, so it belongs in the tool’s own configuration and nowhere else. If one ever ends up somewhere it should not, revoke it from the same screen.

Connecting an assistant is part of Pro, at $8.99 per month, billed annually. The free plan keeps three projects, the browser extension and search, and everything you save there stays yours to export. The full comparison is on the pricing page.