Agents

Bring your own agent.

Keepthru has no assistant of its own. You connect the agent you already use, over MCP, and it works as you: it sees what you see and can do only what you can.

Connect it

Give your agent this address. The first time it calls, your browser opens Keepthru to ask you.

https://app.keepthru.com/mcp

Claude Code

claude mcp add --transport http keepthru https://app.keepthru.com/mcp

Then run /mcp in Claude Code and choose Authenticate.

Codex CLI

codex mcp add keepthru --url https://app.keepthru.com/mcp
codex mcp login keepthru

The second command opens your browser to connect.

Claude on the web and desktop

Add a custom connector in Claude's settings and paste the address above. It signs in the same way.

Anything else that speaks MCP

Gemini CLI, Cursor and others take the same address. Keepthru's sign-in follows the MCP authorization spec.

Things to ask it

"What's Stalled that I'm owed?"
"Log a Check-in on the SSO fix: Sasha says Thursday."
"Draft one Nudge to Pavel for everything he owes me."
"Make a Status write-up for the Halcyon Project."
"Turn the action items from this call into Commitments."

What it can and can't do

Your permissions, no more

Approve each app once, for one Workspace, and choose whether it may make changes or only read. Revoke it at any time in Settings.

Every change on record

What an agent writes is marked as its own and kept in the audit log. Evidence it linked can be detached, and its words go with it.

Other people's words are marked

Text that came from others, such as a Slack message, reaches your agent marked as data to read, never as instructions to follow.

Scripts and the REST API

For scripts and CI, an API token calls the same actions over REST, as you, in one Workspace. The API is described in OpenAPI.

The full guide

Try it on what you owe this week.

The trial is 14 days of the whole product and needs no card. You choose Team or Solo when you subscribe.

Start a free trial