For agents
Your agent does the running around. notanATS does the screening.
Most hiring software assumes a person will sit in it. notanATS is a framework with an agent door: the assistant you already work with can operate it end to end — inside limits that keep the judgment where it belongs.
Connect
Connect your agent in four steps
Works with the agent you already use. You do the setup once — it takes one sitting — and after that your agent does the running around.
ChatGPT
Claude
Google Gemini
PerplexityGet access and sign in
Request access, accept your invitation, and sign in to notanATS. Everything below happens inside your own organisation — nobody outside it can see your positions or your pools.
Turn on agent access
In notanATS, go to Settings → API Keys and turn on “Enable agent access” on your existing key. It is the same key — there is no second key to create, store, or manage.
Give your agent its instructions
Paste the instruction block below into your agent, word for word. When the agent asks for a credential, give it your API key. That is the whole connection — the agent works over the API with your key as its credential.
Or skip the key: if your agent offers a marketplace connection (MCP/OAuth), choose notanATS there and approve the connection.
Your agent runs the loop
From there, your agent can create a position, hand you the calibration link, show you the pool, and advance the people you name. Calibration is always yours, in the notanATS screen — your agent waits while you do it.
Decisions that arrive through a marketplace connection come to you as proposals first — nothing is decided until you approve it.
You are connected to notanATS, a screening framework, on my behalf. Use the notanATS organisation API key I give you as the credential for every call. Treat the key as a secret: do not repeat it back to me or share it with anyone. You may do four things: 1. Create a position from a job description I give you. Before you create it, list my organisation's departments and managers, and ask me which department the position belongs to and who the hiring manager is. When there is more than one to choose from, ask me to choose by name. Never ask me for an ID — the lists give you the ids to use. 2. Return the calibration link to me, then wait. Calibration is my own work, done by me in the notanATS screen, signed in as me. 3. Read the applicant pool for a position: who has been screened, their scores, their standing on each dimension, and any flags on their record. 4. Advance the applicants I name, toward interview booking inside notanATS. You must never do two things: - Never calibrate. Defining the work is mine alone. Do not fill it in, proxy it, or pre-answer it. - Never screen applicants. notanATS itself does the screening. You are a control channel, not a second screener. Work only inside my organisation's positions and pools. When a decision is needed, ask me — do not decide for me.
Paste the whole block. If your agent asks for the key, that is step 3 working as intended.
Using Composio? See the Composio setup guide.
The loop
A position run by an agent looks like this. Note the amber step: it belongs to the manager, and no agent can take it over.
The loop turns around the framework — and then turns again for the next position. One step never moves: calibration belongs to the manager.
The limits
What your agent cannot do
These are not missing features. They are the design.
The definition of the work is yours. An agent that could calibrate would be grading its own homework — the screening would measure the agent's idea of the role, not yours.
Access
One key, scoped to your organisation.
Agent access runs on an organisation API key. The key is shown once when it is created and stored only as a hash. Every call is scoped to your organisation — an agent holding your key can see your positions and your pools, and nothing else's. Keys can be revoked at any time, and calls are rate-limited per key.
The same door serves straightforward integrations: positions, applications, and applicants can be read through the API, always in the rubric-free view.
Why a framework, not a feature?
A screening feature inside a chat tool can only ask questions. A framework holds the whole discipline: the manager's definition, the privacy boundary around the rubric, the scoring, the pool, the applicant feedback, and the fairness monitoring — whoever or whatever operates it. Your agent gets the discipline, not just a question generator.
Get started
Put your agent to work on screening.
Request access, turn on agent access on your key, and connect the agent you already use.
Request access