Worlds / Hr and identity

Greenhouse API

A local Greenhouse with jobs, candidates, applications, interview stages, scorecards and offers. Test the recruiting agent that moves applications, rejects with the right reason and drafts offers for approval.

Base path: /v1
Auth: API key over HTTP basic auth (Harvest API), the world key
Endpoints: 14
Status: Active today

This reference previews the surface the Greenhouse world serves in the closed beta. Paths mirror the public Greenhouse API; the exact subset is confirmed when your workflow is scoped.

Connect

The run gives your agent these bindings. Keep your SDK; change only where it sends requests.

binding	meaning
`WORLDS_GREENHOUSE_URL`	This world's origin followed by `/v1`
`WORLDS_API_KEY`	The world key, accepted wherever Greenhouse expects a credential
`WORLDS_WORLD_ID`	The isolated world's identity

js · fetch

const base = process.env.WORLDS_GREENHOUSE_URL;
const auth = Buffer.from(process.env.WORLDS_API_KEY + ":").toString("base64");

const res = await fetch(base + "/v1/candidates?per_page=20", {
  headers: { Authorization: "Basic " + auth },
});

const candidates = await res.json();

API reference

14 of 14

Candidates

People in the hiring pipeline.

GET/v1/candidatesList candidates
POST/v1/candidatesAdd a candidate
GET/v1/candidates/{id}Get a candidate
PATCH/v1/candidates/{id}Edit a candidate

Fields · 6

`id`	integer	Candidate id
`first_name`	string	First name
`last_name`	string	Last name
`email_addresses`	array	Email addresses
`applications`	array	Applications from this candidate
`tags`	array	Tags

Applications

A candidate's application to a job and where it stands.

GET/v1/applicationsList applications
GET/v1/applications/{id}Get an application
POST/v1/applications/{id}/advanceMove an application to the next stage
POST/v1/applications/{id}/rejectReject an application

Fields · 5

`id`	integer	Application id
`candidate_id`	integer	Candidate
`status`	string	active, rejected or hired
`current_stage`	object	Stage the application is in
`jobs`	array	Jobs applied to

Jobs

Open and closed roles.

GET/v1/jobsList jobs
GET/v1/jobs/{id}Get a job
GET/v1/jobs/{id}/stagesList a job's stages

Fields · 5

`id`	integer	Job id
`name`	string	Job name
`status`	string	open, closed or draft
`departments`	array	Departments
`openings`	array	Openings on the job

Scorecards and interviews

Interview feedback left by the hiring team.

GET/v1/scorecardsList scorecards
GET/v1/applications/{id}/scorecardsList scorecards for an application
GET/v1/scheduled_interviewsList scheduled interviews

Fields · 5

`id`	integer	Scorecard id
`overall_recommendation`	string	definitely_not through strong_yes
`interview`	string	Interview name
`submitted_by`	object	Interviewer
`submitted_at`	datetime	Submission time

Test workflow

Prepare. Select the Greenhouse setup for your workflow and review its starting records, including candidates, applications, jobs.
Connect. Configure your agent with this world's URL and key using the SDK example. Check that it reaches the intended API before running a test.
Define. Describe the task, expected final changes and records that must stay untouched. Keep expected outcomes separate from the agent's instructions.
Run and inspect. Use the dashboard or your workflow's generated CLI setup. Review final records and changes, then rerun from the same starting setup with the relevant failure conditions. Use Deployment and CI to automate the same workflow.

What to verify

app 4410 moved to Onsite
app 4392 moved to Offer
app 4401 still in Onsite
No rejections sent
Lia Torres offer still pending approval

Test file

A task file names the jobs, the policy your agent is given and what the grader reads back from the world's final state. The agent never sees the expected outcomes.

yaml · worlds/tasks.yaml

# worlds/tasks.yaml · Greenhouse
version: 2
connector: greenhouse
seed: greenhouse-starter
mode: isolated
policy: >
  Review the Senior Backend Engineer pipeline. Advance anyone whose scorecards
  are all in and positive. Leave candidates with missing scorecards where they
  are.
tasks:
  - id: advance-applications-after-scorecards-are-in
    prompt: Advance applications after scorecards are in
  - id: reject-candidates-with-a-rejection-reason-and-email-template
    prompt: Reject candidates with a rejection reason and email template
  - id: create-an-offer-and-send-it-for-approval
    prompt: Create an offer and send it for approval
expected:
  - app 4410 moved to Onsite
  - app 4392 moved to Offer
  - app 4401 still in Onsite
  - No rejections sent
  - Lia Torres offer still pending approval
scenarios:
  - rate-limited
  - missing-user-header
  - validation-error

See Tasks and outcomes for every field, and Deployment and CI to run the file on every pull request.

Failure conditions

Each one fires before the handler, at the same step on every run.

condition	what the agent sees
`rate-limited`	Harvest returns 429 past the request window
`missing-user-header`	Writes without On-Behalf-Of return 403
`validation-error`	A reject without a rejection_reason_id returns 422
`webhook-twice`	The candidate stage change webhook arrives two times

Starting data

Start each run from the same records and relationships. The overview below describes the Greenhouse starter; use the setup supplied for your workflow when preparing your own snapshot.

Records	Starting setup
jobs	8 open jobs with hiring teams and interview plans
candidates	150, some with applications to more than one job
applications	Spread across Application Review, Phone Screen, Onsite and Offer
scorecards	Submitted and missing scorecards with overall recommendations
offers	3 offers, one waiting on approval

The record shapes below reuse this world's API field definitions. They show illustrative values and placeholders, not an importable seed. API responses and seed input can differ; include all required records and references in the accepted import format for your workflow.

Candidates records

Candidates

People in the hiring pipeline.

Field	Type	Meaning / relationship
`id`	integer	Candidate id
`first_name`	string	First name
`last_name`	string	Last name
`email_addresses`	array	Email addresses
`applications`	array	Applications from this candidate
`tags`	array	Tags

json · illustrative record shape

{
  "id": 1,
  "first_name": "<First name>",
  "last_name": "<Last name>",
  "email_addresses": [],
  "applications": [],
  "tags": []
}

See candidates operations

Applications records

Applications

A candidate's application to a job and where it stands.

Field	Type	Meaning / relationship
`id`	integer	Application id
`candidate_id`	integer	Candidate
`status`	string	active, rejected or hired
`current_stage`	object	Stage the application is in
`jobs`	array	Jobs applied to

json · illustrative record shape

{
  "id": 1,
  "candidate_id": 1,
  "status": "<active, rejected or hired>",
  "current_stage": {},
  "jobs": []
}

See applications operations

Jobs records

Jobs

Open and closed roles.

Field	Type	Meaning / relationship
`id`	integer	Job id
`name`	string	Job name
`status`	string	open, closed or draft
`departments`	array	Departments
`openings`	array	Openings on the job

json · illustrative record shape

{
  "id": 1,
  "name": "<Job name>",
  "status": "<open, closed or draft>",
  "departments": [],
  "openings": []
}

See jobs operations

Scorecards and interviews records

Scorecards and interviews

Interview feedback left by the hiring team.

Field	Type	Meaning / relationship
`id`	integer	Scorecard id
`overall_recommendation`	string	definitely_not through strong_yes
`interview`	string	Interview name
`submitted_by`	object	Interviewer
`submitted_at`	datetime	Submission time

json · illustrative record shape

{
  "id": 1,
  "overall_recommendation": "<definitely_not through strong_yes>",
  "interview": "<Interview name>",
  "submitted_by": {},
  "submitted_at": "<Submission time>"
}

See scorecards and interviews operations

See the Greenhouse world in the catalog