Connect a repo to Nightjar.
Five steps from “install the GitHub App” to “first agent-shipped PR in front of a human reviewer”. This page is the install walk-through — the fleet overview lives on the product page.
Overview · 01
The doc landing is the install walk-through.
Nightjar is the always-on fleet — a coordinated group of AI engineers that triage, fix, and ship enterprise code around the clock. The fleet runs against your repos the same way it runs against ours: signed PRs, policy gates, human merge. What lives on this page is the path from a clean org to a reviewed PR on the first commit.
Use the product page for what the fleet is, the integrations page for the partners it composes with, and the steps below to actually turn it on. To get on the pilot lane before public availability, the waitlist is the front door; the product page is the longer read.
Getting started · 01
Connect a repo in five steps.
From the org-level GitHub install to the first agent-shipped PR in front of a human reviewer. Each step compounds on the one before — start with a single pilot repo and widen once the review cadence feels right.
Pick the org, accept the permissions Nightjar asks for — read repos, write PR comments, write checks. None of the permissions let us merge or rewrite branches outside an open PR.
One click per org, no service account to provision.
Pick the repos you want Nightjar’s fleet to watch — start with one or two pilot repos before widening. Scope is per-repo, reversible from the GitHub install settings.
Start narrow. Widen once the pilot reviews feel right.
A .nightjar/policy.yml in each repo sets the bars — test, lint, security, coverage, FinOps regressions. Reviewed at PR open, re-run on every push, version-controlled alongside the rest of your config.
Policy-as-code, not knobs in someone else’s dashboard.
Every PR gets a reviewer pass, a security linter pass, and a cost-delta card. The PR is the change-ticket — comments, verdict, and projected spend all live where your engineers already look.
First PR is the signal: PR-01 fires within seconds of open.
If a review turns up blockers, the Fix Shipper picks up the thread, drafts a change in a new commit on the same branch, and hands the diff back to PR-01 for re-review. The human stays in the merge seat.
Failure becomes a draft commit, not a dropped ticket.
The policy gates and the per-agent pause-kill switch are wired in day one — read the security page for the governance details, or the product page for which agent owns each step.
Articles · 02
Read the guide, not the marketing.
Each article walks one install surface end to end — the first PR, the audit trail, the monorepo config — with the design rationale the rest of the docs site flattens out. Read the cards below for the long-form version of the page above, or read the latest shipping notes on the Changelog index.
Pick one repo. Watch the first PR go through.
Pilot scope starts with a single repo under shadow-run — Nightjar files PRs, your reviewers see the trail, and the cost-delta card shows up next to every diff. Production flips on lane-by-lane once your team is comfortable.