Self-hosted and open source. Repositories, credentials, logs, and task history stay on your infrastructure, and the Apache 2.0 codebase is yours to read, audit, and adapt. ProPR Hosted keeps them on a dedicated VM that the ProPR team operates.
ProPR runs next to your codebase from published Docker images, with visible workers, local logs, and the agents you choose. Every external system has one role: GitHub hosts review, and the agent you select receives model requests. The database, task history, queue, repository clones, and settings sit on infrastructure you control.
Verify it yourself: read the source, inspect public PR #1642, or open the live demo.
Each implementation gets its own Git worktree, branch, and Docker workspace. The agent holds a read-only GitHub token, so ProPR makes every branch, commit, push, and pull request. Optional restricted network mode, off by default, starts the container with no network and routes its traffic through a per-run allowlist proxy; every denied host shows on the task timeline. The isolation model.
Every change lands as Git objects and GitHub records: branch, commits, PR conversation, and review scores. The ProPR web UI and CLI show the plan, prompt, logs, model, tokens, and cost. Clone, grep, and diff the history with the tools you already run. The guide to keeping a human approval step on every AI-written change shows the plan approval, PR review and manual merge that produce that record.
A long-running goal works in the same isolated workspace. When the agent implements directly, ProPR opens the draft PR and makes every commit and push at validated checkpoints. Your corrections enter the goal timeline verbatim.
Repository clones, indexes, task files, and logs live in the ProPR stack you run. GitHub access uses scoped permissions through your own App or the ProPR Connect relay; with the relay, your stack holds a revocable relay credential and never the GitHub App private key.
GitHub access, provider credentials, agent subscriptions, and API keys live in your deployment, set by you. Running AI coding agents on your own server lists what the Docker host needs.
The agent you select for a plan or follow-up receives the prompt and relevant context; agent routing sets that per phase. With MCP, your chat host and its model service also receive whatever the conversation returns.
Generated code arrives as branches and pull requests, and your team reviews, checks, and merges it in GitHub.
The desktop app pairs through browser approval and receives an instance token, encrypted with the operating system's credential store. Its renderer runs sandboxed without Node, shell, or file-system access.
ProPR publishes inline review copies of opt-in screenshots and videos to GitHub. Opt-in Connect Plus storage keeps originals in private Cloudflare R2, with metadata that may identify the repository and task. Anyone authorized for the installation can view its originals across all of its repositories.
Plans, task history, logs, usage, and settings stay in your deployment. Each external system receives only what its role requires.
ReceivesRepository access, branches, pull requests, review comments, status updates, completion comments, and inline preview copies.
Kept in ProPRPlans, task history, logs, usage records, worker state, and operational settings.
ReceivesThe prompt and code context for the selected plan, implementation, review, or follow-up.
Kept in ProPRUnrelated repositories, full task history, unused credentials, queue state, and local settings.
Runs hereDocker workloads, storage, database state, repository clones, Git worktrees, logs, and worker capacity.
Kept in ProPRThe application, queue, indexes, isolated runs, and audit trail.
RelaysFull GitHub webhook payloads, cleared after successful delivery. Failed deliveries are cached for replay for up to 24 hours on Plus and 1 hour on the free tier; payload-free delivery statistics (repository, target username, status, timing) are kept 30 days on Plus and 7 days on the free tier. Connect never runs agents.
Stores with PlusPreview originals and metadata when you opt in, including any captured UI or code: 25 GiB per installation, 500 MiB per object, 90-day retention. Privacy covers access and deletion.
ReceivesImage pulls for ProPR and supported agent runtimes.
Kept in ProPRRepositories, task contents, credentials, logs, and review history.
Authenticated MCP gives a chat host scoped access to your instance. Direct MCP connects the chat host to your instance and bypasses Connect. The hosted route sends tool arguments and results through Connect and a managed Cloudflare tunnel; those payloads can include prompts, source code, diffs, and artifacts, and Connect processes them with no end-to-end encryption through the gateway.
Agents, repository clones, execution, and task records stay self-hosted. Results you request also reach the chat host and its model service, under their data policies.
The gateway keeps no application store or replay cache of MCP requests and results. Connect stores only the sign-in records it needs, and the privacy policy describes how it clears them. Cloudflare, your infrastructure, and your chat host apply their own logging and retention. The webhook retention windows above apply only to webhooks. MCP documentation.
Every MCP action checks the grant and your current user and repository permissions. Revoke an app in Connected apps and ProPR denies its new requests; work already accepted finishes. Secrets go through browser sign-in and setup, never through chat.
With ProPR Hosted, repository clones, prompts, task history, logs, settings, and the AI provider credentials you configure live on a dedicated VM that the ProPR team operates. Operators can access that VM to run, secure, back up, and restore the service.
Encrypted application backups hold the database, configuration, and stored credentials. They exclude clones, workspaces, and logs. ProPR Hosted lists each plan's limits; the terms and privacy policy carry the full detail.
The Secure VPS Deployment guide takes the same stack to a hardened server and layers network, identity, and authorization controls in front of the isolated runtime.
Who can trigger agent runs, which repositories are connected, which agents are allowed, which comments start follow-up changes, and which model handles routine versus high-value work. ProPR turns each answer into a setting, next to GitHub comment control, cost visibility, and task history.
Adopting ProPR in a team