● multi-agent dashboard

orbit

Manage multiple AI coding agents simultaneously.
Claude Code, Codex, OpenCode — real-time feed, persistent history, token tracking.

Star on GitHub

Windows 10+ · macOS 13+ · Ubuntu 22.04+ · MIT

see it in action

One window. Every agent.

Orbit — multiple AI sessions running simultaneously

Claude Code Codex OpenCode

multi-agent orchestration

Agents that spawn agents

Orbit includes a built-in MCP server. Any agent can delegate tasks to other providers through standard tool calls — no configuration needed.

● claude-code orchestrator

○ codex refactor API

○ claude write tests

○ opencode update docs

orbit_create_agent orbit_send_message orbit_get_status orbit_cancel_agent

tool call example


  "tool": "orbit_create_agent",
  "input": 
    "provider": "codex",
    "task": "Refactor the user API to use async handlers",
    "cwd": "/home/dev/project"

what it does

Built for the way you actually work.

⊙

Multi-provider

Run Claude Code, Codex, and OpenCode sessions side by side — each with their own models and configs.

⊞

Split panes

View multiple sessions simultaneously. Drag from the sidebar to split into any layout.

◇

Sub-agents monitor

Track spawned sub-agents in a dedicated tab. Click any agent to inspect its full conversation.

◈

Persistent history

SQLite-backed sessions. Close the app, reopen — conversations resume automatically.

Cost & tokens

Per-session token usage, context window %, rate limit bars, and estimated cost in USD.

⎇

Git worktree

Isolate each session in its own branch. Works locally and via SSH remote.

CtrlC

interrupt agent

/model

switch model

/effort

set thinking

@file

pick a file

↑↓

message history