.webp?dpl=dpl_BDbBBE75TxhmYvXB33yHPeYSqjJK)
Agent-operable ATS
The ATS your coding agent can operate — under your approval.
An agent-operable ATS is a hiring system of record that a coding agent like Claude Code or Codex can operate through a CLI or public API to prepare hiring work, while you approve every sensitive action. Yardstick is built this way.
The definition
A hiring system your agent operates — not one that hires for you.
Most “AI ATS” and “AI recruiter” content quietly implies autonomy: software that screens, ranks, or hires on its own. An agent-operable ATS is a hiring system of record that a coding agent can operate through a CLI or public API to prepare hiring work, while a human approves every sensitive action. It is a hiring system of record — roles, candidates, interview plans, scorecards, and decisions — that a coding agent can operate on your behalf to prepare work, with a human approving anything that touches a candidate or a decision.
Agent-operable means the agent can prepare hiring work and hand it to you for approval — draft a job description, build an interview plan, generate questions and a scorecard, summarize a candidate's evidence.
Autonomous would mean the software making and acting on hiring decisions by itself. Yardstick is not that, and does not claim to be. AI prepares the evidence and drafts the decision; people decide.
How your agent connects today. In Yardstick, the agent surface is the yardstick CLI, the public API, and an MCP server, available on every account. You operate the system through your own agent — a coding agent like Claude Code or Codex, or a general assistant like ChatGPT or Claude connected over MCP — which prepares the work and brings it to you to approve.
The framing Yardstick leads with — an agent operating the system via CLI, API, or MCP to prepare work, with humans approving every sensitive action — is how the product actually works.
How it works
Your agent prepares the work. You approve what matters.
You don't chat with Yardstick. You work through an agent— a coding agent like Claude Code, Codex, or Cursor, or a general assistant like ChatGPT or Claude connected to Yardstick's MCP server — and it runs the yardstick CLI, calls the public API, or speaks MCP to do hiring work. Yardstick is the system of record behind it, the same way your repo is the system of record behind the agent that writes your code.
- 1You ask your agentIn plain language, you ask your coding agent for a hiring task — draft this job, build an interview plan, move this candidate forward.
- 2The agent runs the yardstick CLI / public APIYour agent runs the yardstick CLI (or calls the public API) to prepare the work against your live hiring data.
- 3The agent shows you what it preparedThe agent shows you exactly what it prepared and explains precisely what will change.
- 4Approval where it mattersFor certain sensitive changes — like publishing a job or emailing a candidate — Yardstick asks you to approve before it makes them. Routine work, your agent just does. Every change and approval is recorded.
A worked example: you ask your agent to move a candidate into Screening and assign an interviewer. The agent prepares the change in Yardstick and stops.
You: Move Morgan Reed to Screening, assign Alex as interviewer, and add a 30-minute screening task.
Codex: Prepared in Yardstick. Approve before I save the candidate stage change?
- Stage: Applied → Screening
- Interviewer: Alex Carter
- Screening task added
Approval required before saving candidate stage changes.
What needs your approval
Your agent does the work. For a short list of sensitive changes — the ones that affect a candidate, a decision, or your account — Yardstick asks you to approve before applying them. Everything else, it just does.
| Your agent does on its own | Yardstick asks you to approve first |
|---|---|
| Draft a job description | Publishing a job |
| Create an interview plan | Sending a candidate email |
| Generate interview questions and scorecards | Advancing or rejecting a candidate |
| Summarize candidate evidence | Making or communicating a hiring recommendation |
| Prepare scheduling and email drafts | Changing permissions, API tokens, or billing |
The point isn't to slow you down — it's that on the few changes that really matter, Yardstick checks with you first and records it. Everything else, your agent just handles. That's the useful version of an agent in hiring: it moves the work forward, and the decisions that affect a candidate's future still run through you.
This works because the hiring data underneath it is structured. Yardstick is a structured-interview ATS: job-specific plans, standardized questions, and required scorecards mean the agent has real evidence to summarize and draft from, instead of guessing.
Why it matters
Run hiring the way you already run your stack.
If you already run deploys, migrations, and infra through a coding agent under review, an agent-operable ATS lets you run the operational parts of hiring the same way: the agent does the preparation, you approve the decisions. You stay in one operating model instead of context-switching into a heavy ATS UI for every small change.
Founders running a headless stack
You discover and drive most of your tools through Claude Code, Codex, or Cursor, and you want hiring to live in the same place.
Agent-savvy and automation-savvy operators
You want a clear action model and a hard approval boundary, not vague “AI magic.”
Recruiting and hiring leaders
You’ve heard “agent-operable” or “headless ATS” and want to know what it means and whether it’s safe. It is — because the human approves every sensitive action.
Mixed teams — some on agents, some on a UI
Part of your team works primarily through agents; others prefer a software interface. Both operate the same system of record, behind the same approval boundary.
Where Yardstick fits. Yardstick is a focused, interview-led ATS, not a broad AI sourcing or recruitment-marketing suite. If your priority is high-volume AI sourcing and outreach, or recording-first interview intelligence, another tool may fit that specific job better — and the ranked shortlist of AI-hiring ATSes lays out who wins which segment. Yardstick's lane is the team that wants AI woven into a structured hiring workflow under human control, and an ATS their coding agent can actually operate.
Pricing follows the same no-friction idea: Yardstick is pay-as-you-go — you pay for active hiring, not for seats, headcount, or an annual contract — and the public API and agent-operable workflows are included on every account.
If you're new to the broader model, the ATS for humans and agents page covers the concept, and what is Yardstick covers the product. When you're ready to see it on your own hiring, book a call.
FAQ
Common questions about agent-operable and headless ATS.
What is an agent-operable ATS?
An agent-operable ATS is a hiring system of record that a coding agent — like Claude Code or Codex — can operate through a CLI or public API to prepare hiring work, while a human approves every sensitive action. The agent drafts and prepares; the human decides. It is not autonomous hiring.
What is a headless ATS?
Headless means the ATS can be driven through a CLI or public API rather than only through its own screens, so a coding agent or your own scripts can operate it. Yardstick is operable this way via the yardstick CLI and public API, with human approval gates on every sensitive action.
Can my coding agent operate the ATS via CLI?
Yes. The agent surface is the yardstick CLI, the public API, and an MCP server — that's how your agent connects. Your agent prepares the work and brings it to you, and you approve every sensitive action before it happens.
Can an AI agent run hiring for me?
No. The agent prepares the work and explains what will change; you approve before a job is published, a candidate is emailed, a candidate is advanced or rejected, or a recommendation is made. AI prepares evidence and drafts decisions; people decide.
Where do I have to approve?
Before any action that affects a candidate, a decision, or your account: publishing a job, sending a candidate email, advancing or rejecting a candidate, making or communicating a hiring recommendation, and changing permissions, API tokens, or billing. An agent can draft and prepare all of these, but each one waits for your sign-off.
How is this different from the best ATS for AI hiring?
This page explains what an agent-operable ATS is and how it works. For a ranked, side-by-side shortlist of the ATSes that do AI hiring well — and which one fits which kind of team — see the best ATS for AI hiring page.
An ATS your coding agent can operate — with you in control.
See how your agent runs the yardstick CLI to prepare job descriptions, interview plans, scorecards, and candidate summaries — while you approve every sensitive action.