Yardstick MCP server: 176 tools for hiring, in Claude, ChatGPT and Claude Code

Yardstick's hosted MCP server (https://api.yardstick.team/mcp) gives an AI assistant 176 tools for jobs, candidates, applications, interviews, and scorecards. Nothing to install, no API key, and it is on every account, including the free one.

Run your first 3 Jobs free Connect it to your assistant

What it is.

MCP is the open protocol AI assistants use to call tools in other software. Yardstick publishes its hiring tools over that protocol from a server Yardstick hosts, so the assistant you already work in can operate your ATS. You describe the role, the candidate, or the question in your own words; the assistant picks the tools and calls them; the Job, the interview plan, or the report lands in Yardstick where your team can see it.

Hosted. Nothing to install, nothing to run, and no API key to paste into a configuration file. Point your client at the URL.
Browser sign-in. OAuth 2.1 with PKCE, in your browser. Every tool is documented in the tool reference, readable without an account; everything that touches hiring data needs a signed-in Yardstick user.
One list for everyone. Every client sees the same 176 tools, each with a description, its input schema, and a read-only marking where it applies, so your assistant can say what it is about to do before it calls one.
On every account. The MCP server, the public API and the yardstick CLI ship on every account, including the free one. Pricing is pay-as-you-go on open roles.

The MCP server is one of three ways your own agent reaches Yardstick. A coding agent such as Claude Code or Codex can also run the yardstick CLI, and your own code can call the public API. Which ATS your agent can operate compares the three and sets them beside what other ATS vendors publish.

Add Yardstick to your assistant.

The same URL works everywhere. Pick your client.

Claude Code. Run this once, then run /mcp and finish the browser sign-in.

claude mcp add --transport http yardstick https://api.yardstick.team/mcp

Claude, ChatGPT, Perplexity, and other chat apps. Add a custom connector in the app's settings and paste in https://api.yardstick.team/mcp. The app reads the server and lists the tools it found. Step-by-step instructions for each app are in MCP for chat apps.

Editors that use an mcp.json file. Add a server entry with the URL.

{
  "mcpServers": {
    "yardstick": {
      "url": "https://api.yardstick.team/mcp"
    }
  }
}

Cursor. Cursor connects through a local server that the yardstick CLI runs on your machine. The setup is in MCP for developer tools, along with the Yardstick CLI.

Yardstick identifies a connecting app through a Client ID Metadata Document rather than OAuth Dynamic Client Registration. Apps that support the metadata-document flow find this on their own. An app that can only register dynamically stops with an error naming dynamic client registration; that is not a misconfiguration on your side. Use the Yardstick CLI or a local MCP server with that app instead.

What the 176 tools cover.

The tools follow the hiring record: from the Job, through the candidates and their applications, into the interviews and the scorecards, out to the reports. Under each group is the kind of thing you would type to your assistant.

Jobs and job postings

Create and update a Job, write and validate a draft posting, and preview what publishing it needs. Assign a hiring pipeline to a job, and read the pipeline's stages.

“Create a Job for a senior account executive and draft the posting from these notes.”

Candidates and prospects

Read and update people, list Talent CRM prospects and talent pools, add a prospect to a pool, log an activity, preview a prospect import, and create source and referral links for a published posting.

“Which prospects in the product design pool have we not touched in sixty days?”

Applications

List the applications in each stage of a role, read one in full, preview a stage change, and read the decision brief for an application.

“List everyone in the phone screen stage for the account executive role, and read me the decision brief for the top candidate.”

Interviews

Generate an interview plan draft stage by stage: the role outcomes, the competencies, the skills, the rounds, the questions and interviewer directions, and the scorecard anchors for levels 1 to 4. Edit any of it by hand, validate the draft, and preview publishing it. Assign interviewers to a session and generate its scheduling link.

“Draft an interview plan for a customer success manager whose first-year job is renewals on the mid-market book.”

Scorecards and decisions

Submit an interview scorecard with explicit ratings, record a final recommendation, and regenerate the decision brief once the panel's scorecards are in.

“Submit my scorecard for this morning's interview: the ratings are in my notes below.”

Reporting and admin

Search recruiting reporting summaries, list API events and the audit log, manage departments and access grants, author stage automation rules, and preview turning on a built-in agent.

“What changed on the engineering manager role this week, and who changed it?”

The full list, with every tool's arguments, is the tool reference, readable without an account. Publishing a job, advancing or rejecting a candidate, sending a candidate email outside an approved sequence, and the hiring recommendation wait for your approval; publishing a job and stage changes that need sign-off run through a preview your assistant shows you first. The interview plan it drafts is the AI-designed interview guide, with the questions and the scorecard anchors written for the role you described. The built-in agents are the other half of the same category: standing, single-purpose, and hosted by Yardstick, and your assistant can list them and preview turning one on.

Six tools the hosted server does not offer.

Yardstick documents 182 tools. The hosted server publishes 176 of them. The other six need something a hosted connection cannot carry, so they are left off the list rather than listed and left to fail.

File upload

Each reads a file from your own machine (a PDF or DOCX resume, or a Greenhouse export archive), and a hosted connection has no way to read your disk.

yardstick_people_resume_upload_create
yardstick_people_resume_upload_finalize
yardstick_ats_imports_archive_upload_create

Email-provider connection

Each manages the link between Yardstick and an outside email provider, which needs that provider's own browser sign-in.

yardstick_email_sender_oauth_prepare
yardstick_email_sender_connections_refresh_discovery
yardstick_email_sender_connections_disconnect

All six work in the Yardstick CLI, in a local MCP server that the CLI runs on your machine, and in the dashboard. The short version: hosted is for the 176 tools that read and change hiring data; local is for anything that touches your machine.

Your data.

The server reads and writes your organisation's own Yardstick data, acting for whoever is signed in and limited to what that person can already reach in the app. Connecting an assistant does not widen anyone's access.

What leaves Yardstick is whatever the assistant asks for, and it goes to the assistant you connected, so that app's own terms govern what happens to it afterward. A Yardstick administrator can revoke a connection at any time from your account settings, which cuts off its access immediately.

Yardstick's own handling of your data is unchanged by this connection and is covered by the Privacy Policy, the Data Processing Agreement and the Terms of Service.

Where to find it.

Server URL: https://api.yardstick.team/mcp
Registry name: team.yardstick/yardstick on the official MCP registry.
Public repository: github.com/yardstickinc/yardstick-mcp: the README, the registry manifest, and the directory manifests. There is no product source there and nothing to build.
Documentation: MCP documentation and the tool reference.
For AI models: https://portal.yardstick.team/llms-full.txt is the whole documentation set, including the full tool reference, as one plain-text file an assistant can read in a single fetch.

Questions people ask.

Does Yardstick have an MCP server?

Yes. Yardstick's hosted MCP server at https://api.yardstick.team/mcp gives an AI assistant 176 tools for jobs, candidates, applications, interviews, and scorecards. It is live on every account, including the free one, and its registry name is team.yardstick/yardstick.

How do I connect Claude, ChatGPT or Claude Code to Yardstick?

In Claude Code, run: claude mcp add --transport http yardstick https://api.yardstick.team/mcp, then run /mcp and finish the browser sign-in. In Claude, ChatGPT or Perplexity, add a custom connector in the app's settings and paste the same URL. Editors that read an mcp.json file take a server entry with that URL. Cursor connects through a local server, described in the developer-tools documentation.

Do I need an API key?

No. The server is hosted, so there is nothing to install or run, and sign-in is OAuth 2.1 with PKCE in your browser. There is no credential to paste into a configuration file. Every tool is documented in the tool reference, readable without an account; every tool that touches hiring data needs a signed-in Yardstick user.

Can my assistant upload a resume over MCP?

Not over the hosted server. Six of Yardstick's 182 documented tools are left off the hosted list because they read a file from your machine or run an email provider's own sign-in. All six work in the Yardstick CLI, in a local MCP server, and in the dashboard.

What can the assistant change without asking me?

It acts as you, inside what your account can already reach, and tools marked read-only change nothing. Publishing a job, advancing or rejecting a candidate, sending a candidate email outside an approved sequence, and making or communicating a hiring recommendation wait for your approval. A Yardstick administrator can revoke a connection at any time.

Run hiring from the assistant you already use.

Sign up, add https://api.yardstick.team/mcp to Claude, ChatGPT or Claude Code, and ask for the interview plan for your next role. Your first three Jobs are free.

Run your first 3 Jobs free Read the MCP docs