Integrations · ChatGPT (OpenAI)
RichAPI runs a hosted MCP server at mcp.richapi.ai that authenticates with an x-api-key header. We have not verified the in-app ChatGPT connector path end to end and we are not in OpenAI's connector directory, so this page tells you what is confirmed, what is not, and which client to use today if you want this working in the next five minutes.
Last updated September 22, 2026
Create a RichAPI account at app.richapi.ai/signup and copy your API key.
25 free credits, no card.
Confirmed today: call the REST API directly at https://api.richapi.ai/api/v1/{endpoint} with an x-api-key header. Same key, same credits, no client involved.
Confirmed today: the MCP server works in Claude, Cursor, VS Code and Windsurf through the mcp-remote shim. If you want an assistant doing this now, use one of those.
Not confirmed: adding https://mcp.richapi.ai/mcp as a custom connector inside ChatGPT.
The blocker we care about is whether the connector flow accepts a static x-api-key header rather than OAuth.
Not true: RichAPI is not listed in OpenAI's connector directory.
There is nothing to search for and install.
Copy, paste, edit the brackets, run.
Why it matters
ChatGPT is where most people already work, which makes it the integration people ask for first and the one we are least able to promise. A wrong install snippet costs a developer an afternoon. This page would rather be boring and correct.
RichAPI runs a hosted MCP server. It works in several clients. ChatGPT is the one we have not verified, and this page says so up front because a wrong install snippet on a page like this costs a developer an afternoon and costs us the account.
The server is at `https://mcp.richapi.ai/mcp`. It is a standard remote MCP server and it authenticates with a static `x-api-key` header. Hit it without a key and it returns 401, which is the correct behaviour and a reasonable way to confirm it is up. Two things are true about ChatGPT specifically: **We are not in OpenAI's connector directory.** There is no listing to search for, no app to install, no partnership. If another page implies otherwise, that page is wrong. **We have not confirmed the custom connector path end to end.** ChatGPT can add remote MCP servers as custom connectors, with availability depending on plan, workspace admin settings and whether developer mode is on. The part we cannot promise is the auth handshake: our server wants a static header, and connector flows built around OAuth do not always have a place to put one. Until someone here runs that end to end on a real account and writes down what happened, this page is not going to hand you a snippet. the live OpenAPI at https://api.richapi.ai/api/v1/openapi.json will carry the verified steps when there are verified steps.
**The REST API.** No client, no connector, no shim: ```bash curl -X POST https://api.richapi.ai/api/v1/email_finder \ -H "x-api-key: $RICHAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"first_name": "Jane", "last_name": "Doe", "company_domain": "acme.com"}' ``` Same key, same credit pool as everything else. If your goal is data in a pipeline rather than data in a chat window, this was always the shorter path. **Another MCP client.** The server is verified in [Claude](/integrations/claude), [Cursor](/integrations/cursor), [VS Code](/integrations/vs-code) and Windsurf. Same server, same tools, same credits. If what you actually want is "ask for enrichment in prose and get a receipt back", you can have that in about two minutes in any of them. **A workflow tool.** [n8n](/integrations/n8n) and Clay both call the REST API, and both are a better home than a chat window for anything that has to run on a schedule or across a list.
The MCP tool names map one-to-one onto the REST endpoints: - [`email_finder`](/api/email-finder) — a work email from name plus domain, or a LinkedIn URL. - `email_verifier` — deliverability on an address you already have. - `phone_finder` — a mobile from a LinkedIn URL, or name plus company. - `enrich_company` and `enrich_profile` — both take a LinkedIn URL, not a domain. - [`people_search`](/api/people-search) — profiles filtered by industry, headcount, title, location. Every one is a read-only lookup. Nothing in the catalog sends an email, posts anything, or writes to a CRM, which is the property that makes it safe to let an assistant call at all.
Prepaid credits from one pool. Per-call rates are published per endpoint on [/pricing](/pricing); `people_search` is priced per result, so a cost estimate without a result count means nothing. Current tiers are on [/pricing](/pricing). On `email_finder`, `email_verifier` and `phone_finder` a miss costs zero credits: providers ran, none found data, you were not billed. Every other endpoint bills on a 2xx, including an empty one. That scope is deliberate and we would rather repeat it than let you generalise it. Waterfall responses also carry an `execution_log`, an array of provider attempts with per-attempt statuses, so an empty result tells you whether the providers had nothing or something broke.
**Is there a RichAPI app or plugin in ChatGPT?** No. Nothing to install, nothing in the directory. **Will there be?** Registry and directory submissions are on our list. We are not going to put a date on this page. **Can I paste the MCP URL into ChatGPT's connector setup and see what happens?** You can, and if it works we would genuinely like to hear about it. The uncertainty is the header auth, not the server. **Can ChatGPT call the REST API through a custom GPT Action instead?** Action schemas with API-key auth are a documented ChatGPT feature, but we have not tested ours through that path and will not write the steps from memory. **Do I need a credit card?** No. 25 free credits on signup.
25 free credits, no card. If you want this working in a chat window today, connect [Claude](/integrations/claude) or [Cursor](/integrations/cursor) in two minutes. If you want it in a pipeline, the curl above is the whole integration. Every client we have verified is listed on the [integrations hub](/integrations).