Make Claude.ai answer like it knows you
Connect Stateful to Claude.ai over MCP, then prove it works with the Drift Test.
Once Stateful has your context, you can give Claude.ai access to it. After this, Claude answers questions about your work, your people, and your decisions without you pasting anything.
1. Add Stateful as a connector
In your Claude.ai settings, add a custom connector (depending on your plan and version, Claude labels this "Connectors" or "Integrations") and paste this URL:
Claude registers itself and sends you to sign in to Stateful and approve access. On the consent screen, choose the workspace you want Claude to read, and grant:
- Read — so Claude can pull your context
- Write — so Claude can propose facts and edits (these still wait in your inbox; nothing is saved without you)
That's the whole setup — no tokens to copy.
2. Run the Drift Test
Start a fresh Claude conversation and ask, with no other context:
"What am I working on right now?"
If Claude answers with your real projects and priorities, it's working — it pulled them from Stateful behind the scenes. If the answer is generic, nudge it once ("check my Stateful context first"), and confirm your connector is added and your sources have finished their first sync.
3. Correct Claude, and keep the correction
When Claude gets something wrong, correct it in the conversation:
"I'm not the founder — I'm the CTO."
With write access, Claude can propose that fix back to Stateful. It lands in your inbox for approval. Approve it once, and the correction is there for every future conversation — and every other AI you connect.
The rule that makes this work
The one behavior that matters: the AI should fetch your context before answering anything about you, using your real question as the search. That's what separates a grounded answer from a guess. If you connect other MCP-capable tools, the same URL works everywhere — see Use the same context in every tool.