Now in private beta
Connect Linear and GitHub. Orbital picks up issues, implements the code, tests, fixes, and merges PRs — autonomously. Your 4-person startup ships like a team of 20.
We use Orbital to build Orbital. Every feature in this product was implemented, tested, and merged by an agent.
Link your Linear workspace and GitHub repo. Add a playbook.toml that defines your agent workflow.
Create issues in Linear. Describe what you want built. Mark as ready. Orbital picks them up and starts working.
The agent implements in a branch, runs your test suite, fixes failures, opens a PR on GitHub, and merges when ready.
Tests fail. The agent reads the output, patches the code, tries again. Review rejects. Back through the fix loop. No limit on iterations. Each pass builds on the last — context accumulates, the code improves.
The system keeps going until the issue is resolved. You check in on a PR, not a process.
Something goes wrong mid-run? The agent picks up where it left off. No orphaned branches, no lost work, no manual cleanup.
Multiple issues run simultaneously, each in its own isolated workspace. No conflicts. They all merge cleanly.
Tests fail? The agent reads the output, understands what broke, fixes it, and tries again. No human needed in the loop.
Define exactly how your agent works in a simple config file. What it implements, what it tests, how it reviews. Your process, automated.
Orbital runs your actual tests. bun test, pytest, cargo test — whatever you use. No separate validation layer.
Triage issues before you leave. Come back to merged PRs. Your team focuses on architecture and product — the agents handle the tickets.
Playbooks define what the agent does for each issue. Chain agent steps, shell scripts, and decision gates. Branch on success or failure. Issue context is injected automatically via variables.
This is the actual playbook we use to build Orbital.
[[steps]]
id = "implement"
type = "agent"
prompt = """
Implement ticket $TICKET_ID: $TICKET_TITLE
$TICKET_BODY
"""
[[steps]]
id = "test"
type = "script"
command = "bun run test"
on_success = "review"
on_failure = "fix"
[[steps]]
id = "fix"
type = "agent"
prompt = "Fix the failing tests..."
next = "test"
[[steps]]
id = "review"
type = "agent"
prompt = "Review the implementation..."
on_failure = "fix"Bring your own API key, or let us handle the tokens. No per-run limits.
Write issues in Linear. Wake up to merged PRs. We're onboarding startups now.
Request accessNo credit card required · Free tier available · Cancel anytime