Orbit
▎ what it does · 20+ features in alpha

Everything the office runs on.

Desks, job board, gates, ledger. Every feature shipping in the Orbit alpha, and the panels you can open right here.

▎ try the panels

Open the real surfaces.

These are the app's own panels, not mockups. Apollo is the receptionist you talk to; the rest are the rooms it can walk into.

  1. 01

    Multi-project prompt queue

    Stack work across every project. Run sequentially or in parallel. Walk away. Come back to merged PRs, not a stuck terminal.

  2. 02

    Any agent at any desk

    Claude Code, Codex, Gemini, Grok, DeepSeek, Kimi, and anything OpenCode can run. Same queue, same context, same vault. Route each job to whichever model is best for it.

  3. 03

    Apollo · embedded chat

    A built-in AI chat with full Orbit tool access and MCP. Ask it to ship — it has the queue, the vault, the pipeline, the codebase.

  4. 04

    Pipeline · unified feed

    Every project's work in one live feed — approval gates, QA badges, dependency chaining. Deploy on green. Replay on red.

  5. 05

    Encrypted vault

    Project keys, rotation policy, deploy tokens — encrypted at rest, surfaced when needed.

  6. 06

    Completion log

    Full audit trail of every shipped prompt — inputs, outputs, diffs, tokens, model. Replayable.

  7. 07

    Memtrace

    A live graph of your code — symbols, call sites, dependencies, history — handed to every agent so it stops re-reading the codebase. The efficiency screen shows what that saves: context, tokens, hours.

  8. 08

    Mobile orchestration

    Queue prompts, approve sessions, monitor the pipeline, and dispatch builds to iOS, Android, Windows, and Mac. All from your phone.

  9. 09

    Image Studio

    AI image, video, and icon generation built in. For when the next thing is an Open Graph card, not a commit.

  10. 10

    Sandbox mode

    Isolated branch testing — run risky changes in a clean copy, then promote or discard with one click.

  11. 11

    Daily summary

    AI narrative of what shipped, timeline view, pattern detection across projects.

  12. 12

    Boards

    Agents write real documents, not chat: plans, status pages, decision records, per project or across all of them. You read them in the sidebar. The next agent reads them back.

  13. 13

    QA system

    Automatic post-completion quality checks — types, tests, lints, screenshots.

  14. 14

    Coordinator mode

    For builds too big for one prompt. Each subtask gets its own git worktree and its own branch off a shared base commit, and the results land on an integration branch — your main is never touched until you say so. You approve the plan and its projected cost before anything runs. At the merge, every claim a subtask made is checked against the actual diff, and each divergence surfaces on its own with investigate, accept, or send-back.

  15. 15

    Dev servers, deploys, Dispatch

    Start and stop local servers from the queue. Dispatch a build to iOS, Android, Windows, Mac, or the web: TestFlight, Play Store, Vercel, Railway, with reviewed release notes. From the desktop or your phone.

  16. 16

    Glossary + Learning Atlas

    300+ terms of the modern AI dev landscape, each on its own page, and a Learning Atlas that puts them in reading order. Public, on the docs site.

    Open the glossary →
  17. 17

    Bridge

    A private channel where your Claude and their GPT can share project context directly. No copy-pasting. No switching tabs.

  18. 18

    Agent wiki

    Every project gets a wiki its agents write for each other. A finding made on one run, with evidence, reaches the next run in its briefing, so the same guard, the same flaky test and the same dead end are learned once. In alpha now.

  19. 19

    Linux projects

    Run Ubuntu-native projects through Orbit while the dashboard stays on Windows — files, commands, agents, servers and builds all route into Linux. Guided setup checks WSL2, the helper, and your Linux agents before anything runs.

  20. 20

    Usage meters, every provider

    Claude, Fable, Codex, Grok, AGY: session and weekly quotas with resets, in one panel. Alerts name the provider and the window that tripped, not just a percentage.

  21. 21

    Benchmark mode

    Run the same prompt across models on your own project and compare what each one shipped, what it cost, and whether QA passed. Find out which model is best at any task in your workflow.