# Use Kepler with Claude (MCP)

> Kepler, the AI recruitment CRM for agencies, has an MCP server. Connect Claude, ChatGPT or Codex to search candidates and update records.

_Collection: AI & Automation (ai-automation) — Kepler Help Center. Canonical: https://keplercrm.com/support/articles/use-kepler-with-claude/_

Kepler works with Claude. Kepler is an AI recruitment CRM for agencies, and it runs a hosted MCP server, so Claude can search your candidates and update your records with the permissions you grant. The same connection works from Claude Code, ChatGPT and Codex.

## Which recruitment CRM works with Claude?

Kepler is a recruitment CRM that connects to Claude through the Model Context Protocol (MCP). You add Kepler to Claude as a custom connector, sign in with your Kepler account, and Claude can then work with your candidates, contacts, companies, jobs and pipeline.

Kepler's MCP server is hosted by Kepler. There is nothing to install and no local server to run.

## Does Kepler have an MCP server?

Yes. Kepler's MCP server is at `https://app.keplercrm.com/api/mcp`. It uses Streamable HTTP and OAuth sign-in, so you connect by signing in to Kepler rather than by copying a key.

Kepler gives setup steps for four AI clients:

  - **Claude** on the web and in the desktop app

  - **Claude Code**

  - **ChatGPT**

  - **Codex**

## What can Claude do in Kepler?

With read access, Claude can search candidates in plain English across profiles, CVs, notes and call summaries. It can also look up jobs, companies, contacts and applications, count and break down records, read activity history, check your task list, and read the match results already stored for a job.

With write access, Claude can:

  - create candidates, contacts, companies and jobs

  - update standard and custom fields

  - add and edit notes

  - create and update tasks

  - move candidates between pipeline stages and record outcomes

  - add people to lists and remove them

Things you can ask Claude once Kepler is connected:

  - "Using Kepler, who have we spoken to about senior finance roles in Manchester this year?"

  - "Log a note on this candidate from my call and set a follow-up task for Thursday."

  - "How many open jobs do I own, and which have no applicants yet?"

## How do I connect Claude to Kepler?

**Before you start:** AI client access is switched on for each workspace. If you do not see **AI clients** under **Settings → Developers**, email [support@keplercrm.com](mailto:support@keplercrm.com) and ask for it to be turned on for your workspace.

### Claude on the web or desktop

  - In Claude, open **Settings → Connectors** and choose **Add custom connector**.

  - Paste `https://app.keplercrm.com/api/mcp` and select **Connect**.

  - Sign in to Kepler when asked. Choose the workspace and the permissions you want Claude to have, then approve.

### Claude Code

```
claude mcp add --transport http kepler https://app.keplercrm.com/api/mcp
```

Claude Code opens the Kepler sign-in the first time it uses the connection.

### ChatGPT and Codex

ChatGPT adds Kepler as a custom MCP app, and Codex adds it with one command. See [Use Kepler with ChatGPT (MCP)](/support/articles/use-kepler-with-chatgpt/) for the steps.

> **TIP:** To check the connection, ask "Using Kepler, how many open jobs do I own?". If the client answers from memory, say "use the Kepler tools" once.

## What stays in my control?

  - **You choose the permissions.** At sign-in you tick what the client may do: read, write records, tasks, notes, pipeline and lists. Read alone is a good place to start.

  - **Your Kepler role still applies.** A connection can never do more than you can do in Kepler yourself.

  - **One workspace per connection.** You pick the workspace when you approve the connection.

  - **No delete tools.** An AI client connected through MCP cannot delete records.

  - **No silent overwrites.** If a colleague changed a record after Claude read it, Kepler refuses the update and Claude has to read the record again.

  - **Every call is logged, and you can revoke access.** Revoke a connection in **Settings → Developers → AI clients** and the client's next call fails.

## Does Kepler have an open API as well?

Yes. Kepler has a public REST API and outbound webhooks for your own tools and integrations. API keys are scoped. Each key can make 300 reads and 60 writes per minute. Webhook deliveries are signed and retried, with a delivery log. Create keys and webhooks in **Settings → Developers**.

Use MCP when you want an AI assistant you already pay for, such as Claude or ChatGPT, to work with your Kepler data. Use the API when you are building an integration or connecting a back-office system.

## Do I still need Kepler's own AI Chat?

AI Chat is Kepler's built-in assistant. It searches your CVs, notes, emails and call transcripts in plain English and asks for your approval before it changes anything. AI Chat is included on every plan, monthly and annual, with no add-on or upgrade needed.

MCP is an extra way in. Use AI Chat inside Kepler, and use Claude or another AI client when you are already working there.