openchamber/openchamber

OpenChamber – AI‑assisted coding workspace

What it is

An open‑source application that lets you run coding agents (AI models that write or modify code) and manage the whole development workflow.
Works on desktop (macOS, Windows, Linux), as a web/PWA, inside VS Code, and on iOS/Android.

Key capabilities

Feature	What you can do
Session Goals	Define a concrete outcome for a session. The agent keeps working until the goal is met, it gets blocked, or a user‑set limit is reached – even if you close the app.
Multi‑run & Fusion	Run the same task on up to five models in parallel, compare results, or merge the best parts into a new session.
Changes Walkthrough	Turn a large diff into an AI‑guided step‑by‑step tour that explains each edit.
Preview	Inspect a running web app side‑by‑side with the chat. Click an element to send the agent a screenshot, styles, and errors, so it can reason about the UI directly.
GitHub integration	Start a session from an issue or PR, feed failed checks or review comments to the agent, and update/merge the PR from within OpenChamber.
Cross‑device continuity	Open the same projects and sessions on any supported surface; the state is synced so you can pick up work anywhere.
Private Relay	One‑time QR‑code pairing creates an end‑to‑end encrypted tunnel (or use LAN/VPN, Cloudflare, Ngrok, SSH) without exposing public ports.
Dashboard & scheduling	See status of all sessions, token usage, costs, and organise them in folders. Schedule prompts to run on cron‑like intervals, optionally with Session Goals.

How you use it

Install – download a desktop binary, add the VS Code extension, or run the Node‑based CLI (curl … | bash).
Start a session – pick a model via the OpenCode API, give it a prompt or a GitHub issue, and optionally set a goal.
Interact – chat with the agent, send code snippets, request previews, or let it run automatically.
Review – use the built‑in diff viewer, Changes Walkthrough, or Fusion to evaluate results.
Ship – merge PRs, push commits, or export the final code directly from the UI.

Why it matters

Bridges the gap between raw LLM output and a full development workflow: version control, review, testing, and deployment are all in one place.
Keeps the AI “on track” with goals and automatic continuation, reducing the need for constant manual prompting.
Supports multiple models side‑by‑side, making it easy to benchmark or combine their strengths.

License: MIT – you can freely use, modify, and redistribute the software.

All information above is taken directly from the repository’s README.