Leonar's MCP server connects Claude, ChatGPT and any MCP agent to your recruiting data. Ask in plain language and your agent sources candidates, enriches their contact details, updates your CRM and enrolls them in your sequences.
Quick answer: Leonar MCP lets Claude, ChatGPT, GitHub Copilot and any MCP-compatible agent act on your recruiting data through structured tools. In plain language, your agent understands the role, searches and refines candidates, enriches their contacts, updates your CRM and ATS, drafts a sequence, and enrolls contacts into an active one, without screen scraping.
The same tool layer powers Copilot, the AI assistant built into every Leonar screen (⌘J). Meet Copilot Leonar
Works on any Leonar workspace with API access, including Professional, Enterprise and eligible trials.
MCP is an open standard created by Anthropic that lets AI assistants connect securely to external tools and data sources. Instead of copying data between your CRM and your AI tool, MCP creates a direct, secure bridge between them. Your AI agent reads and writes to Leonar as if it were a native user.
Think of MCP as a universal adapter for AI. Before MCP, every integration between an AI assistant and a business tool required custom code, API wrappers, and ongoing maintenance. MCP replaces all of that with a single, standardized protocol. AI tools like Claude, ChatGPT, and Cursor already support it natively, which means connecting them to Leonar takes minutes instead of weeks of engineering work.
For recruiting teams, this is a game changer. Instead of switching between your CRM, your sourcing tools, and your AI assistant, you can work entirely through natural language. Ask your AI to find candidates, update a pipeline, or draft an outreach message, and it executes the action directly inside Leonar. No copy-pasting, no manual data entry, no context lost between tools.
MCP turns your AI assistant into a recruiting co-pilot. Here are the scenarios where teams see the biggest impact.
Instead of building complex boolean filters manually, describe the candidate you need in plain English. Your AI agent searches across your full candidate database in Leonar, applies the right filters, and returns a ranked shortlist. A search that used to take 15 minutes now takes seconds.
After a round of interviews, tell your AI assistant to move candidates between stages, add interview notes, or flag top performers. The agent updates Leonar in real time, so your pipeline always reflects the latest status without anyone typing into spreadsheets or clicking through forms.
Ask your AI agent to draft personalized emails from each candidate's profile, experience and the role you are filling. It pulls the context from Leonar and writes messages that read personal, not templated. It prepares the sequence as a draft; you activate it, and only then can contacts be enrolled into it.
Instead of scrolling every thread, ask your AI to read and summarize recent candidate conversations across email, WhatsApp and SMS, flag who is waiting on a reply, and create the follow-up tasks. The agent reads your unified inbox in Leonar and turns it into a short, actionable brief.
Each step maps to real tools on the MCP server. Ask in plain language, and your agent runs it against your workspace. Here is the request to try and what changes in Leonar.
Point your agent at a live project and it reads the sourcing brief: the role, the must-have skills, the seniority and the location you are hiring for. Result in Leonar: the agent works from the exact criteria your recruiters set, not a guess.
"Read the sourcing context for the Senior Backend Engineer project"
Describe the profile in plain language, then narrow the list with follow-up filters until the shortlist is right. Result in Leonar: a ranked shortlist you can open and action, with the search resumable by its id.
"Find senior React developers in Paris with 5+ years of experience"
"Now keep only the ones who worked at a SaaS company"
Ask the agent to find a work email, personal email or phone number for a shortlist. Enrichment runs on your workspace credits and reports its status. Result in Leonar: verified details are added to the record where they are found, since not every lookup returns a hit.
"Find work emails for these 10 candidates"
Tell the agent to log a note, create or update a contact, or move a candidate along the pipeline. Result in Leonar: your CRM and pipeline records reflect the change immediately, with no manual form-filling.
"Log today's call with Marie Laurent and move her to the interview stage"
Ask the agent to draft a multi-step follow-up sequence with the copy and cadence you want. Result in Leonar: a new sequence is created as a draft for you to review and activate yourself. Creating this new draft sequence does not send any messages.
"Draft a 3-step email follow-up for candidates who went quiet"
Once a sequence is active, ask the agent to enroll the right contacts and personalize up to five variables each. Result in Leonar: contacts join your active sequence and start its email, WhatsApp and SMS steps.
"Enroll these 12 contacts into my active Onsite sequence"
MCP gives your agent the actions. Whether it runs once or all day is decided by the environment you run the agent in, not by the protocol. The Leonar MCP server is stateless and answers one request at a time.
Scheduling
Your agent platform decides when the agent runs: on your prompt, on a timer, or on an event. Leonar responds per request and does not schedule the work itself.
Memory
What the agent already sourced and where it stopped lives in your agent environment. Leonar holds the workspace data; the agent holds its own working context between runs.
Usage limits
Leonar enforces rate limits, plan capacity and billing checks on every call. Your agent runtime adds its own budget and model caps on top.
Reporting back
Configure the agent to report to the recruiter after each run: what it searched, enriched, updated or prepared. Documented once, the loop repeats on its own.
This is a documented configuration pattern. Measured day-and-night usage numbers are reserved for a separate client write-up.
Prefer Leonar to run the agent for you? Meet the built-in AI agents
For Claude Desktop and Claude.ai, add the hosted connector URL and authorize your workspace through Leonar OAuth. Local clients can use an API key fallback.
In Claude Desktop or Claude.ai, open Settings / Customize > Connectors, choose Add custom connector, and enter https://app.leonar.app/api/mcp. Do not use the old SSE URL.
Claude discovers Leonar OAuth from the endpoint, redirects you to Leonar, and asks which workspace the connector may access. This flow does not require a Leonar API key.
Claude Code, Cursor, Codex, and generic MCP clients can connect with Direct HTTP Bearer auth or the local @leonar/mcp NPX bridge using a leo_... API key.
The Leonar MCP server is compatible with any AI assistant that supports the Model Context Protocol.
The same enterprise-grade security as our REST API protects every MCP connection.
No. For Claude Desktop and Claude.ai, setup means adding https://app.leonar.app/api/mcp as a custom connector, signing in to Leonar, and authorizing one workspace. A Leonar API key is only needed for local or non-OAuth MCP clients.
Leonar's MCP server works with MCP clients that support Streamable HTTP or a local stdio bridge. Claude Desktop and Claude.ai should use the remote OAuth connector. Claude Code, Cursor, Codex, and generic clients can use Direct HTTP or the @leonar/mcp NPX bridge.
Yes. Claude remote connectors use Leonar OAuth tokens scoped to a single workspace. API-key clients use the same leo_... keys and scopes as the REST API. All requests are rate limited, checked against billing access, and bound to workspace membership.
You control what your agent can reach. Remote OAuth connectors are scoped to a single workspace. API keys carry scoped permissions, so you can limit an agent to specific resources and to read or write access. The agent can only perform the actions its scopes allow.
The Leonar REST API is for developers building integrations directly. MCP exposes curated recruiting tools and resources to AI assistants over https://app.leonar.app/api/mcp, with OAuth for Claude remote connectors or Bearer API keys for local clients.
Every action your agent takes through MCP is recorded in your Leonar activity log, so you can review exactly what was queried or changed and undo it. Whether the agent pauses for your confirmation before acting is decided by your agent environment, not forced by the MCP server. The sequence design helps here: your agent can draft a new sequence for review, and can only enroll contacts into a sequence you have already activated.
Practical guides for sourcing with Claude, connecting MCP to your recruiting stack, and using AI agents to grow your agency.
Use AI agents like Claude and ChatGPT to find, target, and win new clients for your staffing agency. Includes prompts, connected workflows, and templates.
Use Leonar's open API and MCP protocol to connect Claude, ChatGPT, or any AI agent to your recruiting CRM. Step-by-step guide with real use cases.
Connect Claude or ChatGPT to Leonar and let it source, enrich, update your CRM and enroll candidates while your team keeps control.