What makes Greenhouse easy to automate
Greenhouse exposes nearly the whole recruiting record through the Harvest API: jobs, candidates, applications, stages, scorecards, interviews and offers. Webhooks fire on the events recruiting teams care about, such as new applications, stage changes, offers and hires. The Job Board API publishes open roles and accepts applications on your own careers site. Zapier’s Greenhouse app covers common triggers without code.
The main recent change is authentication. Harvest v3 replaced v1 and v2, which were retired in 2026. It uses OAuth 2.0, with client credentials for a company’s own integrations and authorization code for partner apps, and every request is tied to a Greenhouse user. That makes a dedicated integration user with tight permissions the right setup.
The AI layer
Recruiting has a lot of writing and coordination, and AI is useful for both. It is also a regulated area, so where AI sits matters:
- Coordination. Scheduling interviews across busy panels. See our interview scheduling guide.
- Writing. Job descriptions, candidate emails and interview summaries, all reviewed by a person.
- Reading. Summarizing applications against written requirements so recruiters review faster. Our responsible resume screening guide explains how to keep people in charge of decisions.
We do not build systems that automatically reject or rank candidates. EEOC guidance holds employers responsible for discriminatory outcomes from AI tools, and NYC Local Law 144 requires bias audits and notices when automated tools substantially assist hiring decisions for NYC roles.
Limits to plan around
- API migration. Anything still pointing at Harvest v1 or v2 has stopped working. Migrate to v3, including its pagination.
- Permissions. Harvest v3 credentials carry scopes you manage in Greenhouse, and list endpoints need a Site Admin as the authorizing user. Grant only what the workflow uses, especially for offers and private notes.
- Rate limits. Harvest enforces request limits per time window. Bulk syncs need paging and backoff.
- Attribution. v3 actions show as the integration user (or the user named in the token request). Name that user clearly so the audit trail makes sense.
- Candidate data. Applications contain personal data. Minimize what goes to models and follow your retention policy.
How Lumeor typically connects it
We create a dedicated Greenhouse integration user and OAuth client, receive webhooks in n8n, call Harvest v3 with HTTP Request nodes, and use Google Calendar or Outlook for scheduling. AI steps draft and summarize; recruiters approve in Greenhouse or Slack. When a candidate is hired, the workflow hands off to the HRIS, often BambooHR or Workday, and starts onboarding.