O

Orbi

Self-hosted delivery runner driven by GitHub Issues. Pi writes code in an isolated worktree; a separate session reviews it before merge

agent_platform Website ↗ Website ↗ Verified 2026-11-17

📊 Scores

Coding Ability
5.0
Value
8.0
Flexibility
4.5
China Access
5.0

🚀 Quick Start

Install Steps

  1. 1 Read https://docs.orbi.build/ and the repo README. Do not use referral query parameters
  2. 2 Prepare git, curl, authenticated GitHub CLI ≥2.94 and Linux systemd/macOS launchd; Node ≥22.19/npm is needed when Pi is missing
  3. 3 Follow official getting-started: uv can provision Python 3.14 and installer can add Pi; then configure orbi.toml and a model connection
  4. 4 Label a GitHub Issue ai-ready. For a release, open a separate release Issue and label it ai-release
Register / Get Account ↗
🇨🇳 China Access Note: A README.zh-CN.md and docs/zh exist, but access from mainland China was not tested. The board depends on GitHub; self-hosting uses your model endpoint, while Cloud clones the repo onto Orbi's machines

💲 Pricing

Plan Price Limits
Self-hosted $0 No platform fee. Bring your own model (OpenAI-compatible API, a local endpoint, or a Codex login inside Pi) Official →
Cloud trial $0 First 3 successful merges need no credit card; failed deliveries do not count. Drafting an Issue does not use one of the three Official →
Cloud Solo $29/mo 400M tokens/month, 1 repo at a time; pauses on exhaustion. Regular annual total $290; limited promotions separate Official →
Cloud Pro $79/mo 1.2B tokens/month, up to 5 repos; pauses on exhaustion. Regular annual total $790; limited promotions separate Official →

Rechecked 2026-11-17: regular Cloud Solo $29/mo (400M tokens, 1 repo), Pro $79/mo (1.2B tokens, 5 repos); first 3 merged PRs free, exhausted allowances pause new deliveries with no per-token overage. Limited founding offers are not a guaranteed price for everyone; regular annual totals are $290/$790. Dual AGPL-3.0/vendor SUL 1.0 licensing requires following the chosen terms. Self-hosting has no platform fee, but GitHub/remote models may still receive code; Cloud clones authorized repos onto vendor machines. Historical editorial scores retained, not newly measured or re-rated; not a fresh performance or cost benchmark.

Orbi

Free software/tier listed; check pricing scope and signup terms

Website ↗

✨ Key Features

Code Completion —
Chat —
Agent Mode ✅
Multi-file Edit ✅
MCP Support —
Sub-agent —
Multi-Agent —
agent_teams —
task_management ✅
open_source ✅
self_hosted ✅
issue_tracking ✅
independent_review ✅
github_only ✅
Context Window 取决于所配模型
supported_runtimes Pi
supported_os Linux, macOS
ide_base Agent 协作平台

✅ Pros

  • +The board is GitHub Issues and labels; the repo has no second database or queue
  • +The Pi implementation session is separate from review; public sample PR
  • +Self-hosting has no platform fee, and the model can be any OpenAI-compatible endpoint

⚠️ Cons

  • −GitHub only. The installer accepts Linux (systemd) and macOS (launchd); the README says the macOS scheduler is not yet verified on real hardware
  • −Requires GitHub CLI ≥2.94, git, curl and Linux systemd/macOS launchd; uv can provision Python 3.14 and installer can install Pi, but not Node ≥22.19
  • −A release is not a side effect of merge: someone must open a release Issue and apply ai-release. Cloud clones authorized repos onto Orbi's machines
  • −This was source/documentation/pricing review, not installation or an end-to-end delivery on a third-party repo; independent review does not guarantee catching all errors

🎯 Best For

Repos whose work already lives in GitHub Issues and should come back as a PRTeams that want implementation and review in separate sessions, with the record left on GitHubPeople who can bring their own model and self-host on Linux

📝 Changelog

2026-10-16 Added from public repo v0.7.0, the install script, and the website. Scores are CodePick's; the install was not run

📚 Related Guides

🔗 Related Tools

💬 Comments

Last Verified: 2026-11-17 Next Review: 2026-12-17 Confidence: low Report Error →