Run Claude Code, Codex and other coding agents on your own server. Every task runs in an isolated workspace and lands as a pull request you review and merge in GitHub.
Every ProPR release ships through ProPR's own plan, run and review loop, in a public repository.
Seven projects went from issues to merged pull requests through ProPR, among them a game from an empty repository, products built from written plans and codebases with years of history. All seven are live today.
Game
Claude Cope
458merged pull requests, 412 written by agents
Web app
mcptest.io
200agent-written pull requests merged over sixteen months
Commercial product
dig.vin
185merged pull requests between 17 July and 11 September 2026
Business product
A messaging-assistant platform
770+merged pull requests, 650+ written by agents
All 7 case studies →New to GitHub? Your first AI pull request, start to finish →
ProPR 0.9.1 · 1 November 2026 · Release notes
Hand an agent a standing objective. Follow its progress, correct it mid-run, and review the result as a draft PR.
Run a saved prompt on a schedule or from CI. Each run writes a report, and you choose whether it becomes tasks.
Switch agents to restricted mode: no network, only a per-run allowlist proxy. Task agents get read-only tokens; ProPR does every push.
ProPR for Linux and macOS: saved instances and accounts, native menus, and guided local setup on Linux.
Run plans, goals, tasks, reviews and automations from an MCP client. An administrator enables it, and you choose which repositories and scopes it gets.
Install the web app, get push notifications per category, and set quiet hours.
Manage AI agents like human engineers: through issues, pull requests and code review. Label a GitHub issue, get a pull request back, and steer it with ordinary review comments.
AI labelTriggers the agent run.A model label such as llm-claude-fable51 picks the agent. Put several model labels on one issue and ProPR opens one PR per model, so you compare their work side by side.
Plan issues, steer goals, inspect diffs, and review and fix pull requests from Claude, ChatGPT or any MCP client.
Each chat client connects through browser consent, reaches only the repositories you pick, and never exceeds the scope ceiling an admin sets. Agents and execution stay on your infrastructure.
Explore MCP from chat → Set up MCP. Compare access and privacy for each route.
Large features get a plan before code. Finished pull requests get a second opinion from a different agent. Both land in GitHub.
ProPR turns a prompt and your repository into review-sized tasks with acceptance criteria. Refine the plan in conversation, approve it, and ProPR files the GitHub issues.
Explore planning
Comment /review and ProPR checks the diff against the issue, then posts numbered merge blockers, suggestions, and a score. It never touches the code. Reply /fix with the IDs you keep and the fixes land on the same branch.
Every run gets its own container and Git worktree, so parallel agents never collide and your main checkout never changes.
Run many tasks on one repository at once, each on its own branch.
Every task keeps its prompt, live logs, commits and per-call model cost.
Claude Code, Codex, Antigravity CLI, Mistral Vibe or OpenCode, on your existing AI subscription or API key.
Epic mode runs a multi-issue plan in order and merges each green pull request before the next task starts.
ProPR is free, open source, and self-hosted. Repositories, task history, logs, and credentials stay on your server, behind a documented security model. Bring the AI subscriptions or API keys you already pay for; ProPR never marks up your tokens. Prefer a managed instance? ProPR Hosted runs a dedicated one for you.
Runs on a Linux Docker host.
npm install -g propr-cli
propr setupStraight answers on scope, safety, billing and setup.
ProPR is the engineering process around coding agents. Plan the change, run it in an isolated workspace, get a pull request, and refine it inside GitHub. Claude Code, Codex, and the other agents write the code. ProPR gives them one process your team already trusts.
Sign in to your own ProPR instance with GitHub. Self-hosting is free, and ProPR Hosted runs a dedicated instance for you. Try the product at demo.propr.dev, then start your own with the quickstart.
A terminal session needs you at the keyboard, approving each command and edit. ProPR moves your approval to the pull request. Label the issue, walk away, review the PR. It works like delegating to another engineer.
Yes. Agents hold a read-only GitHub token, and ProPR makes every branch, commit, and push itself. Optional restricted network mode limits each run to an allowlist of hosts. Every result arrives as a pull request you review, merge, or discard. Trust and security has the full model.
No. ProPR runs on the subscriptions you already pay for, including Claude Pro and ChatGPT Plus, or on direct API keys. Use one provider or several. ProPR never marks up tokens.
No. ProPR runs on GitHub pull requests, review comments, status checks, and comment commands.
ProPR fixes it. With automatic CI follow-up on, a failing check starts a fix from the check's output, and the correction lands as a follow-up commit on the same branch. Or paste the failure into a PR comment and ProPR takes it from there.
A Linux Docker host, GitHub access to your repositories, and the agent subscriptions or API keys you already have. Plans, tasks, and logs stay on your server. The quickstart takes you from install to the first agent PR.
Step-by-step answers live in the guides. For architecture and security, read how ProPR works →
Live IDE agents, hosted agent platforms, and review bots, side by side with ProPR.