For AI agents
A CRM your agent can work in, and you can check.
One MCP server and one agent key. Your agent finds people, reads their history, logs what happened and moves deals; each change carries its name and an Undo.
Claude Code, Cursor, or any MCP client that sends a key
claude mcp add --transport http tieloom https://tieloom.com/api/v1/mcp \
--header "Authorization: Bearer tlk_your_agent_key"{
"mcpServers": {
"tieloom": {
"url": "https://tieloom.com/api/v1/mcp",
"headers": {
"Authorization": "Bearer tlk_your_agent_key"
}
}
}
}Seven tools: search people, read a record, today's list, the stages, log what happened, add a follow-up, move a deal. Make an agent key under Connections; a key made for a product can only send entries.
Seven tools.
Named and described the way your agent sees them, so it picks the right one without a prompt from you.
search_peopleFind people by name, phone, email or company, optionally in one stage.get_personOne person's record: details, stage, timeline (newest first) and tasks.todayWhat needs someone today: open tasks due today or earlier, urgent first, and the latest entries.list_stagesThe workspace's pipeline stages, in order.log_entryAdd a call, message, email, meeting, note, form, booking or request to a person's timeline, found or made by phone or email. The Autopilot then reads it.add_taskAdd a follow-up for a person, optionally due on a date.move_stageMove a person to another stage, with a one line reason the team will see.
Access you can see and take back.
- Only an owner gives access
- An agent key is made by a workspace owner and can be removed in one click. A key made for a product, like Damaruka's, can only send entries; it cannot open the CRM.
- Every change is signed
- What an agent does lands in the Autopilot log under the key's name, beside the reason it gave, so the office sees which agent did what and why.
- Every change can be undone
- A follow-up or a stage move made by an agent has an Undo, like the Autopilot's own. Anyone who can edit can reverse it.
Claude Desktop, too.
Claude Desktop reaches remote servers through mcp-remote. Add this to its config file, with your agent key in place of the example, and restart it.
Settings, Developer, Edit Config
{
"mcpServers": {
"tieloom": {
"command": "npx",
"args": [
"mcp-remote",
"https://tieloom.com/api/v1/mcp",
"--header",
"Authorization:${TIELOOM_AUTH}"
],
"env": {
"TIELOOM_AUTH": "Bearer tlk_your_agent_key"
}
}
}
}Questions people ask
Which agents work with Tieloom?
Any client that speaks the Model Context Protocol over HTTP and can send a header: Claude Code and Cursor directly, Claude Desktop through mcp-remote, and agents you build yourself.
What can an agent see?
The workspace its key belongs to: people, their timelines, tasks and stages. Nothing from another workspace, and nothing about the team's sign ins or keys.
Can an agent send messages to my customers?
No. Tieloom has no tool that sends anything to a customer. An agent can log, add follow-ups and move deals.
How is this different from the API?
The API is for a product or a script that sends entries in. MCP lets an agent read the CRM and act in it, with every change signed and undoable.
What does it cost?
Nothing while Tieloom is new.
Hand your agent the CRM.
Make an agent key under Connections, paste one line into Claude Code or Cursor, and ask it what needs someone today.
Free while Tieloom is new. No card.