Skip to main content
Free 3-min audit: how tech-mature is your agency? Start the audit

MCP server for recruiting

Ask Claude or ChatGPT about your candidates, clients and jobs

Leonar has an MCP server for recruiting: your AI assistant works directly in your Leonar data. Connect it once, then ask in plain English.

  • Claude, ChatGPT or any MCP client
  • More than 100 tools
  • Included in Professional and the trial
Book a demo Start a trial

Already on Leonar? Connect your AI in 3 steps

Used by 400+ high-growth recruiting teams and agencies globally

  • Robert Half logo
  • Lemon.io logo
  • Glorium Technologies logo
  • Talent Solvers logo
  • 1871 Advisors logo
  • BCI logo

Watch, 31 seconds

Ask in Leonar, in Claude or in ChatGPT

Leo is the AI copilot built into Leonar. The MCP server brings the same Leonar actions to Claude and ChatGPT. One request, the same result, wherever you work.

Read what the video shows

A recruiter asks Leo, the AI copilot in Leonar: "Who hasn't replied on the Finance Director job?"

Follow-ups: four candidates are still waiting. Pipeline: "Move Hannah to Client interview." Her stage changes and the call note is saved on her record. Client updates: "Draft an update for the client." The draft is ready for review, with 12 contacted, 5 interviews and 2 with the client. New business: "Who decides at Northgate?" Leo lists the decision makers at the company.

Contacts, companies, projects, pipeline, sequences, deals, notes, tasks, tags, conversations, sourcing and applications. More than 100 actions, one ask away.

Leo prepares, you approve. An approval card asks to enroll the four candidates in a follow-up sequence, and nothing is sent until the recruiter clicks Approve.

Ask in Leonar, in Claude or in ChatGPT: "Log my call with Northgate, move the deal to Proposal, remind me Friday." The call note is saved, the deal moves to Proposal and the task is created. The same Leonar actions, through our MCP server.

Just ask Leo. Leo is built into Leonar. Try it free at leonar.app, with a 14-day free trial and no credit card.

Once connected

What you can ask your AI

Requests from a normal agency day. Your AI picks the right Leonar tools and answers in the chat.

  • Follow-up

    “Who hasn't replied on the Data Lead project?”

    Reads the project's sequence and lists who is still waiting, with the step they reached.

  • Pipeline

    “Move Sarah Martin to Client interview and log a note from my call.”

    Changes her stage on the project and saves the note on her record.

  • Talent pool

    “Find financial controllers in our database who live near Lyon.”

    Searches the contacts already in Leonar by job title, location and keywords.

  • Client update

    “Where are we on the Data Lead job? Draft a short update for the hiring manager.”

    Reads the pipeline and notes, then drafts the update for you to edit and send.

  • Outreach

    “Draft a 3-step email follow-up for candidates who went quiet.”

    Creates the sequence as a draft in Leonar. You review it and activate it yourself.

  • Business development

    “Find the decision makers at Acme for our Finance Director search.”

    Returns the right people to approach at that company.

Example requests. What your AI can do follows your Leonar role and your workspace modules.

Setup

Connected in three steps

No code for Claude or ChatGPT. You need a Leonar workspace on the Professional plan or a trial.

  1. Step 1: Copy the server URL

    https://app.leonar.app/api/mcp

    Paste it in Claude or ChatGPT as a custom connector.

  2. Step 2: Sign in to Leonar

    Your AI app opens the Leonar sign-in page. You pick the one workspace it may use.

  3. Step 3: Ask

    Type your request in the chat. Your AI calls Leonar and shows you the result.

Using Claude Code, Codex or Cursor? They connect with a Leonar API key instead of a sign-in. The setup guide further down covers both.

Control

What stays under your control

Your AI works with your access, in the workspace you chose.

  • One workspace per connection

    You pick the workspace when you sign in. The connection cannot reach another one.

  • Your account, your plan

    Your AI acts under your Leonar account. Read-only viewers cannot connect.

  • Approvals in your AI app

    Your AI app can ask you before each tool call. Keep that on for anything that writes, enriches or deletes.

  • Outreach starts as a draft

    A sequence your AI creates is saved as a draft. It can only add people to a sequence that is already active.

Included in the Professional plan and in the trial

The MCP server and the REST API come with Professional, next to the ATS, CRM, sequences and inbox. Try it on your own data during the trial.

From

$149

per user per month, annual prepaid

Book a demo Start a trial
See every price and billing term

Guide: MCP servers for recruitment agencies

What an MCP server is, which ATS connects to ChatGPT or Claude, and how to use one safely in an agency.

Contents
  1. 01What is an MCP server?
  2. 02Which ATS connects?
  3. 03Day-to-day use cases
  4. 04Security and permissions
  5. 05Copilot or your own AI
  6. 06Connect each client
  7. 07Getting started

The basics

What is an MCP server for recruiting?

MCP stands for Model Context Protocol. It is an open standard, published by Anthropic in November 2024, that lets an AI assistant use outside tools in a structured way.

An MCP server for recruiting is the part your ATS or CRM provides. It tells the assistant which actions exist, such as "search contacts" or "move a candidate to a stage". When the assistant asks, it runs them on your data. Your AI stops working from copies you paste into the chat. It reads the live record.

For a recruitment agency, this means the AI you already pay for can answer questions about your jobs, candidates and clients. It can also keep your pipeline up to date while you talk to it.

Compatibility

Which ATS connects to ChatGPT or Claude?

Leonar does. Leonar is an ATS and CRM for recruitment agencies. Its MCP server works with ChatGPT, Claude (Claude Desktop, Claude.ai and Claude Code), Codex, Cursor and any other MCP client. It is included in the Professional plan and in the trial.

What makes the difference is what your AI can do once connected. With Leonar, it does more than read. It moves a candidate to a new stage, adds a note, creates a task, prepares a sequence and enriches a contact.

To compare an ATS MCP, ask three questions. What can your AI do once connected? Can a team without a developer set it up? What stays under your control? With Leonar, the answer is on this page: more than 100 tools, three steps, and your approvals.

Use cases

How a recruitment agency uses an MCP server, day to day

Start the day from your pipeline

Ask which candidates are waiting on you, which sequences got replies and who sits at the interview stage. Your AI reads the projects, pipeline stages and conversations in Leonar. Then it gives you a short list to work through.

Prepare a call or a client update

Before a call, ask for a summary of a candidate: current job, notes from past calls, open projects. Before a client meeting, ask for the status of their job, with the number of candidates at each stage. Your AI drafts the update. You edit it and send it yourself.

Search your own talent pool

Describe the profile in plain words. Your AI searches the contacts already in Leonar by job title, location and keywords. When you agree, it adds the right people to the project.

Keep records clean after a busy day

Say what happened: who moved stage, what the candidate said, which follow-up to plan. Your AI updates the stage, saves the note and creates the task. Feedback from the hiring manager lands on the right record instead of staying in your inbox.

Find people to call at a client company

Ask your AI to find the decision makers at a target company for a given project, or to look for new profiles in Leonar Source, our database of 870M+ profiles.

Prepare outreach

Ask for a three-step email sequence for a job. Leonar saves it as a draft. You read it, adjust it and activate it in Leonar. Your AI can then add contacts to that active sequence, with email, SMS and WhatsApp steps.

For a full sourcing example, step by step, read how AI sourcing with Claude and Leonar works. If ChatGPT is your main tool, start with our guide to ChatGPT for recruiting.

Security

Security and permissions when your AI works in your ATS

Sign-in through OAuth

Claude Desktop, Claude.ai and ChatGPT connect through OAuth, the sign-in standard behind buttons like "Sign in with Google". You log in to Leonar on a browser page and pick one workspace. Your AI app then receives access to that workspace only. You never paste a password or a key into the chat.

API keys with scopes

Claude Code, Codex and Cursor use a Leonar API key that starts with leo_. You create it in Leonar's settings and choose its scopes. A key can be read-only, or limited to a few kinds of records.

Your plan and your role apply

MCP access comes with the Professional plan and the trial. Read-only viewers cannot connect, and every request is checked against your workspace membership.

Approval works in two places

In your AI app, you decide whether the assistant asks before each tool call. In Leonar, a new sequence always starts as a draft, and contacts can only join a sequence that is already active. Enrichment uses your workspace credits, so keep approval on for it too.

What your AI reads from Leonar goes through your AI provider, under its terms, so choose an AI plan that fits your candidate and client data.

Two ways to work

Built-in copilot or your own AI: which one should you use?

Leonar gives you two ways to work with AI on the same data. Leo is the AI assistant built into Leonar. The MCP server lets the AI you already use, such as Claude or ChatGPT, work on your Leonar data from its own app. Both draw on the same catalog of tools.

  • Where you use it

    Leo, built into Leonar
    Inside Leonar, next to the record you are working on
    Your own AI through MCP
    In Claude, ChatGPT or another MCP client

  • Setup

    Leo, built into Leonar
    Already on for every account
    Your own AI through MCP
    Copy the URL, sign in, choose a workspace

  • AI model

    Leo, built into Leonar
    Included with Leonar, or on your ChatGPT or Grok subscription, or your API key
    Your own AI through MCP
    The model and subscription you already pay for

  • Approvals

    Leo, built into Leonar
    Leonar asks you before sends, sequence enrollments, enrichment and deletions
    Your own AI through MCP
    Set in your AI app; new sequences stay drafts in Leonar

  • Best for

    Leo, built into Leonar
    Quick help without leaving Leonar
    Your own AI through MCP
    Longer work that mixes Leonar with your other tools

Many agencies will use both: Leo for quick help on a record, their own AI for longer sessions that mix Leonar with email, documents or other connectors.

Setup guide

How to connect Claude, ChatGPT or another MCP client to Leonar

Every client uses the same server URL, https://app.leonar.app/api/mcp. Want the in-app assistant instead? Read more about Leo, the AI recruiting copilot.

  1. Claude Desktop and Claude.ai

    Open Settings, then Connectors, and choose Add custom connector. Name it Leonar and paste the server URL. Click Connect, sign in to Leonar and choose the workspace. On Claude Team and Enterprise plans, an owner may need to add the connector for the organization first.

  2. ChatGPT

    On chatgpt.com, open Plugins, click the add button and create a custom MCP server. Name: Leonar. Connection: paste the same URL. ChatGPT opens the Leonar sign-in page, where you sign in and choose the workspace. Then install the plugin and type @ in a chat to call it. On ChatGPT Business, Enterprise and Edu, an admin or owner creates the server for the team. On Plus and Pro, access and write actions depend on what OpenAI opens to your plan.

  3. Claude Code, Codex, Cursor and other MCP clients

    Create a Leonar API key, then add the server with that key. In Claude Code, one command does it: claude mcp add --transport http leonar https://app.leonar.app/api/mcp --header "Authorization: Bearer leo_your_key"

Clients that only run local commands can use Leonar's small bridge package, @leonar/mcp. Developers who would rather write their own integration can use the Leonar recruiting API instead. For a wider view of agents, MCP and APIs in an agency stack, read how to connect AI agents to your recruiting stack.

Good to know

Getting started with your AI in an agency

Your AI works on request

You ask, it acts. For a daily routine, set it up in your AI app.

One connection per workspace

If your agency has several workspaces, connect each one.

AI sourcing on top

AI sourcing is activated at setup. Its tools then appear in your AI, with access to Leonar Source.

Treat it like a new colleague

Read what it proposes before you approve, especially at first. If you want to see it on your own jobs first, compare the plans or book a demo with the team.

FAQ

Questions about the Leonar MCP server

Short answers on clients, plans and control. Ask the team for anything else.

Book a demo

Which ATS connects to ChatGPT or Claude?

Leonar, an ATS and CRM for recruitment agencies, connects to ChatGPT, Claude and any MCP client through its MCP server. It is included in the Professional plan and the trial. Your AI does more than read: it updates your records, your pipeline and your tasks.

Do I need technical skills to connect Claude or ChatGPT?

No. You paste one URL in your AI app, sign in to Leonar and pick a workspace. An API key is only needed for tools like Claude Code, Codex or Cursor.

Which Leonar plan includes the MCP server?

The Professional plan, and the trial. Read-only viewers cannot connect. Prices and billing terms are on the pricing page.

Can my AI send emails or delete records without asking me?

Not if you keep approvals on in your AI app, which is where you set them. Leonar adds its own guardrails: sequences your AI creates start as drafts, and contacts can only join a sequence that is already active. With an API key, you can also leave out the write and delete scopes.

Is the MCP server the same as Leo, Leonar's copilot?

No. Leo is built into Leonar. The MCP server lets your own AI, such as Claude or ChatGPT, use the same tools from its own app. You can use one or both.

How is MCP different from the Leonar API?

The REST API is for developers who build an integration in code. The MCP server is for AI assistants: it describes each tool so your AI can choose and use it from a chat. Both come with the same plan.

Go further with AI in your agency

Put your AI to work on your own pipeline

Book a demo to see Claude or ChatGPT working in Leonar, or start a trial and connect your own AI.

Book a demo Start a trial