Lever MCP: what exists and how to connect AI tools
The short answer
Lever has no official MCP server. Neither Lever’s help centre, its API documentation nor Employ, Lever’s parent company, describes one as of 1 October 2026. The official route to Lever data is the Lever API: a Super Admin creates API keys for internal workflows, and OAuth is reserved for registered partner integrations. Community MCP servers built on the Lever API exist, but Lever doesn’t support them, and an API key carries the access of the whole key rather than of one user. Lever’s own AI, such as the AI Interview Companion, runs inside Lever.
Every Lever detail comes from Lever’s own documentation, read on 1 October 2026. Beatview is not affiliated with Lever.
Lever and AI assistants at a glance
| What the documentation says | |
|---|---|
| Official MCP server | None documented by Lever or Employ |
| API access | API keys for customers’ internal workflows; OAuth only for registered partner integrations |
| Who creates keys | Super Admins, from the Integrations and API page in Settings |
| Key permissions | Chosen per endpoint, read or write, when the key is created, and can’t be changed afterwards |
| Confidential data | Excluded from a key by default unless you choose to include it |
| Rate limits | 10 requests a second per key, with bursts up to 20; Lever says these defaults are not guaranteed |
| Lever’s own AI | AI Interview Companion and AI Screening Companion (add-ons you purchase), Talent Fit and other features inside Lever |
| Community MCP servers | Exist and use a Lever API key; not endorsed or supported by Lever |
Connecting an assistant through the Lever API
Lever has no official server, so any MCP connection runs on an API key. These steps come from Lever’s own API documentation and help centre.
- 01
Step 01
Agree it with your admin
An API key carries many privileges and isn’t tied to one person, so agree what an assistant may see and change before anyone builds anything. - 02
Step 02
Create a narrow key
A Super Admin creates the key under Integrations and API in Settings, ticking only the endpoints needed, read-only where possible. Leave confidential data excluded unless it’s required. - 03
Step 03
Store it safely
Lever shows the key only once, and its endpoints can’t be edited later; create a new key if your needs change. - 04
Step 04
Stay within the limits
Expect about 10 requests a second per key, with bursts to 20, and retry with exponential backoff when Lever returns a 429.
Lever API limits that affect AI tools
Limits to know
- No official MCP server means no per-user sign-in: anything built on an API key acts with that key’s access.
- Only Super Admins can create API keys, and a key’s endpoints can’t be changed once it exists.
- OAuth, which would allow per-user access, is limited to registered partner integrations.
- Rate limits are about 10 requests a second per key, and Lever says they aren’t guaranteed.
Controls Lever gives you
- Confidential postings, opportunities and requisitions are excluded from a key unless you turn them on.
- Anything created through a key’s write permissions shows the key’s name in Lever.
- Keys can be limited to specific endpoints, read or write.
- Lever’s own AI features run inside Lever and need no outside assistant.
Lever stays your ATS. Beatview brings screening to your assistant.
Beatview’s MCP server connects Claude, ChatGPT, Gemini and Cursor to Beatview itself, with each person signing in to their own Beatview account. Your assistant can open roles, compare applicants by their resume and AI interview scores and send interview invitations, while Lever stays your system of record.
One address
Add https://platform.beatview.ai/mcp to Claude, ChatGPT, Gemini or Cursor. It is included on every Beatview plan, including Free.
Evidence, not just records
Beatview returns each applicant’s resume match score with reasons and a structured AI interview assessed answer by answer.
You stay in charge
The assistant confirms with you before moving a candidate or sending an interview invitation, and recordings never leave Beatview.
Try with Beatview connected
- “Who are the strongest applicants for the Product Designer role, by interview score?”
- “Compare the top three candidates who finished the interview, strongest first.”
- “Who needs action today across all my open roles?”
Lever MCP: frequently asked questions
Still deciding?
Bring a live vacancy and we’ll walk through where automation ends and recruiter review begins.
Not an official one. As of 1 October 2026, Lever’s help centre, its API documentation and Employ, its parent company, describe no Lever MCP server or official connector for Claude, ChatGPT, Copilot or Gemini.
Not with an official connector. Teams can build their own connection on the Lever API with a key a Super Admin creates, or use a community MCP server, which Lever doesn’t support. Either way the connection acts with the key’s access, not each user’s.
Lever allows a steady 10 requests a second per API key, with bursts up to 20 when possible. It says these defaults aren’t guaranteed and recommends retrying with exponential backoff.
Only Super Admins with permission to manage API credentials and integrations, from the Integrations and API page in Settings. The key is shown once, and its endpoints can’t be changed after it’s created.
Yes. Lever offers an AI Interview Companion and AI Screening Companion as add-ons you purchase, plus Talent Fit and other AI features, all running inside Lever.
Yes. Lever stays your ATS and Beatview handles screening and AI interviews alongside it. Beatview’s MCP server lets your assistant work with screened and interviewed candidates on every Beatview plan, including Free.
Where these details come from
Lever’s own documentation, read on 1 October 2026. MCP servers change quickly; check the source before you rely on a detail.
- Lever API documentation
- Lever Help: generating and using API credentials
- Lever Help: AI Companion
- Lever AI features
Beatview publishes this guide and runs its own MCP server, which works alongside Lever. Where Lever does not document something, we say so rather than guess. Spotted a change? Tell us.