Connect an AI tool to Pipelined
In this guide
What this lets you do
Pipelined exposes your recruiting data through the open Model Context Protocol (MCP). Once you connect an AI tool, you can have natural-language conversations with it about your pipeline:
- "List my open roles at Stripe."
- "Add Jane Doe (linkedin.com/in/jane) as a candidate for the Backend Engineer role."
- "Summarize the most recent transcript on candidate X and append your summary to their notes."
- "Move the Stripe Senior Designer role to hold."
The AI can read all of your Pipelined data and create or update candidates, companies, roles, and notes. It cannot delete records — destructive operations stay in the web interface only. See the Privacy Policy for the full data-handling details.
Before you start
You need:
- A Pipelined account (sign in at pipelined.net first)
- An account with the AI tool you want to connect (Claude.ai, Cowork, ChatGPT, etc.)
- The Pipelined MCP server URL — you'll paste this into the AI tool's connector form:
Connect Claude.ai (web)
VerifiedUse this if you chat with Claude in your browser at claude.ai.
- Open Settings → Connectors in claude.ai.
- Click Add custom connector.
- Paste the URL
https://api.pipelined.net/mcpin the Remote MCP Server URL field. - Click Add. A browser tab opens to Pipelined.
- Sign in with the Google account tied to your Pipelined account (skip if already signed in).
- You'll see a consent screen showing which app is requesting access. Click Allow.
- The tab closes automatically. Back in claude.ai, the connector shows as Connected.
Test it: in any conversation, ask "Using my Pipelined connector, list my companies." Claude should respond with the companies in your account.
Connect Claude Cowork (desktop)
Should workCowork is Anthropic's desktop agentic AI for knowledge work. It uses the same connector form as claude.ai.
- Open Claude Cowork on your Mac or Windows machine.
- Open Settings → Connectors (or wherever the connector list lives in your version).
- Click Add custom connector.
- Paste
https://api.pipelined.net/mcp. - Click Add. Your default browser opens to Pipelined.
- Sign in if prompted, then click Allow on the consent screen.
- Switch back to Cowork. The connector should now appear as Connected.
If your Cowork version uses a different label for the connector area (e.g., "Plugins" or "Tools"), the underlying flow is identical.
Connect ChatGPT
VerifiedChatGPT supports MCP via custom apps, but the option is hidden behind developer mode. You'll need to enable that first.
- Open ChatGPT and go to Settings → Apps.
- Click Advanced settings.
- Toggle on Developer mode. (One-time setup — you won't have to repeat this for future connectors.)
- Go back to the Apps screen and click Add app.
- Paste the URL
https://api.pipelined.net/mcpand submit. - ChatGPT opens a sign-in window. Sign in with your Pipelined Google account.
- On the consent screen, click Allow.
- The window closes. The connector appears in your Apps list as Connected.
Test it: in any conversation, ask "Using my Pipelined app, list my companies." ChatGPT should call the connector and respond with your companies.
Connect other MCP clients
Any MCP client that supports remote (HTTP-based) servers with OAuth 2.1 authentication can connect to Pipelined. The only thing you need to provide the client is the URL https://api.pipelined.net/mcp — the client handles discovery (via /.well-known/oauth-authorization-server), dynamic client registration (RFC 7591), and the OAuth dance automatically.
If your MCP client offers manual fields for OAuth Client ID and Secret, leave them blank — Pipelined supports Dynamic Client Registration, so the client will register itself.
Disconnect or revoke access
You can revoke any AI tool's access to your data at any time:
- Sign in to pipelined.net/profile.
- Scroll to Connected applications.
- Find the AI tool you want to disconnect and click Disconnect.
- Confirm the action.
Common issues
"This URL doesn't seem to be reachable"
Make sure you used https://api.pipelined.net/mcp (HTTPS, with the /mcp path) — not pipelined.net alone. Most MCP clients require an HTTPS URL; HTTP isn't accepted.
"You don't have any tools" (or new tools missing after we added them)
AI clients cache the list of tools at the start of each session. If you connected before a tool was added, the AI won't see the new tools until you disconnect and reconnect. The fix: open Connected applications in your Pipelined profile, click Disconnect, then re-add the connector in the AI tool.
"Authorization failed" after clicking Allow
This usually means an underlying credential was rejected at the very last step. Try:
- Disconnect the connector on the AI tool's side, then re-add it.
- Make sure you signed in with the same Google account you use for Pipelined — if you have multiple Google accounts, the OAuth screen might pick the wrong one.
- If the issue persists, contact support with the timestamp of your attempt — server logs help diagnose.
"AI says it can't add candidates / write data"
If you've connected before but aren't seeing write tools (add, update), the AI is likely working from a cached tool list. Disconnect + reconnect refreshes the list.
"I can't find the connector form in my AI tool"
Custom connectors are typically a paid-plan feature on consumer AI products. Check your subscription tier; on free plans, MCP/connectors may not be available.
Need help?
Email support@pipelined.net with a description of what you're trying to do and any error messages or screenshots. Include the timestamp (in your local time) when the issue happened — that helps us trace through server logs.