Skip to main content

Connecting Recruiterflow to ChatGPT via MCP

Learn how to set up the integration, what you can do with it, and how to get the most out of it.

Written by Amogh Balikai

What is Recruiterflow MCP?

Recruiterflow's MCP (Model Context Protocol) integration lets you connect your Recruiterflow account directly to AI assistants like ChatGPT. Once connected, you can search your candidate database, pull up job details, review pipelines, and read candidate activity, all from inside ChatGPT, without switching tabs or logging into Recruiterflow separately.

Think of it as giving ChatGPT a live window into your Recruiterflow account — the same window Recruiterflow already gives Claude.


What ChatGPT Can See Today

The integration covers most of what you'd look up manually in Recruiterflow:

  • Candidates — name, location, current title and company, tags, custom fields, current pipeline stage, jobs attached, and who owns or added the record

  • Jobs — name, visibility (external, internal, unlisted, confidential, draft, closed, or closed confidential), client company, and status

  • Pipeline — the full stage-by-stage breakdown for a job, including timestamps and disqualification reasons

  • Activities — notes, calls, meetings (summary only, not the full transcript), and emails logged against a candidate (job and company-level activity are not yet included)

  • Reference data — team members, pipeline stages, and client companies, so ChatGPT can resolve names like "Sarah" or "Acme Capital" to the right record

The integration is read-only — ChatGPT can query this data but cannot create, edit, delete, or merge anything in your account. Think of it as a reporting layer, not an editing one.

This is the same data set the Claude MCP integration exposes today — ChatGPT and Claude are reading from the same Recruiterflow connector, just through each assistant's own connection flow.


How to Connect

Quickest way — from the ChatGPT plugin directory: Open ChatGPT, go to Plugins, search for "Recruiterflow," and click Add to your account. Recruiterflow is now listed as an officially approved plugin, so most users won't need Developer Mode or the manual setup below at all.

Next, Click Allow on the consent screen
​

Note: Authentication uses OAuth 2.1. Recruiterflow never stores your ChatGPT credentials, and ChatGPT never stores your Recruiterflow password. You're authorising a secure, read-only connection between the two.

Using it in a conversation: Select the Recruiterflow app for a specific message (it only applies to that message, not the whole conversation), or @mention it again whenever you want ChatGPT to pull fresh data or take a new action. You can combine Recruiterflow with other apps in the same prompt.


Example Prompts to Try

Once connected, here are some things you can ask ChatGPT:

Look things up, without opening Recruiterflow:

  • "Find candidate Tony Stark"

  • "Show me candidates in New York or Boston with 'React' in their tags"

  • "Who's currently in the Interview stage of the Backend Engineer job?"

  • "List all open jobs for Acme Capital"
    ​

Ask for the analysis directly:

  • "Break down every candidate on this job by lead owner and stage - [job name]"

  • "Rank disqualification reasons on this job by frequency - [job name]"

  • "How many candidates does each recruiter currently own across every open job?"

Connect the output to other tools:

  • "Turn this job's pipeline breakdown into a table I can drop into a client report"

  • "Summarize this candidate's full activity history into a one-paragraph brief before my call"

Name search matches partial names too — "Tony" will find "Tony Stark," so you don't need the full name to get a hit.


Data & Privacy

  • The integration is read-only. ChatGPT can retrieve data from Recruiterflow but cannot create, edit, or delete anything.

  • Access is scoped to your account. ChatGPT can only see data that your Recruiterflow user account has permission to access.

  • You can disconnect the integration at any time from ChatGPT Settings → Plugins


Did this answer your question?