GTM data for Claude Code, Cursor, and Codex.Ask your agent for a list. Get a list.
One MCP server gives your agent 100+ searches and enrichments across 50+ providers. Waterfalls bill only when they find something, and when a list outgrows the chat, it lands in a sheet your team can open.
claude mcp add --transport http pipe0 https://api.pipe0.com/v1/mcpOne prompt, start to finish.
Your agent finds the right search, runs the waterfall, and writes the result where your team can see it. No scraping, no keys to juggle.
>Find heads of RevOps at Series B fintechs in Berlin, get their work emails, and put them somewhere my team can see.
- run_search_oneshot people:profiles:crustdata@3→ 25 people
- run_pipes_oneshot person:workemail:waterfall@1→ 23 work emails found · 2 misses, not billed
- create_sheet Berlin fintech RevOps→ sheet created
- run_effects rows:add:run@1→ 25 rows written
Done. 23 of 25 have a work email. They're in Berlin fintech RevOps, ready for your team.
More found, through the agent.
Same 20 people, asked through each tool's own agent interface. Count the numbers, then check the bill.
20 records, one run each, September 2026. Test it on your own list. Method and data
When the list outgrows the chat, it gets a sheet.
Data comes back in the call while that's enough. When a list needs a schedule, a webhook, or a teammate's eyes, the agent writes it to a sheet. Same records, nothing rebuilt.
In the call
Records come straight back to your agent, ready for its next step.
// records, back in the call
{ name: "Lena Brandt",
company_domain: "heliolabs.io",
work_email: "lena@heliolabs.io" }
{ name: "Jonas Weber",
company_domain: "payfold.de",
work_email: "jonas@payfold.de" }In a sheet
Up to 2M rows, on a schedule, open to your team and to the agent.
| Name | Domain | Work email |
|---|---|---|
| Lena Brandt | heliolabs.io | lena@heliolabs.io |
| Jonas Weber | payfold.de | jonas@payfold.de |
| Mira Koch | clearsum.com | mira@clearsum.com |
| Tim Ahrens | ledgerly.io | — |
Built for an agent to get right.
Typed blocks that never shift under it, and errors it can read and fix on its own.
Schemas before calls
The agent reads every search and pipe's inputs and outputs before it runs anything, so it builds the payload instead of guessing it.
Checked before billed
Every run is validated first. A wrong key or an unknown filter value comes back as a named error with the closest real value, and nothing is billed.
Asks before it spends
Tools that spend credits or change data confirm on the first call. The same engine, prices, and limits as the API.
filters.locations: "Berlin, DE"
Not a known value. Nothing ran, nothing billed.
closest: "Berlin, Germany"
retry → 25 people
Connect in one command.
Sign in with pipe0 in the browser. No API key to paste, no provider accounts to open.
claude mcp add --transport http pipe0 https://api.pipe0.com/v1/mcpThen run /mcp and choose Authenticate.mcpServers in .cursor/mcp.json, then sign in from Settings, MCP.https://api.pipe0.com/v1/mcpPay per result. Misses are free.
MCP calls are billed like API calls. No subscription required, no second meter for your agent.
Mobile number
average cost per result
- pipe0
- 14¢
- Typical
- 25–100¢
Work email
average cost per result
- pipe0
- 4.5¢
- Typical
- 8–30¢
Actions
CRM writes, Slack messages, and sequence enrollments on your own connections. No second meter that grows with every step.
$3*per 100k actions
* On plans above $300/mo
Give your agent the data.
Connect the MCP server, sign in once, and ask for your first list. New accounts get free credits.