Frequently asked questions about how ProPR works.
ProPR is self-hosted, Apache 2.0, and built around GitHub. Short answers here; the documentation has the exact wiring.
Jump to a topic
- Getting started8 questions
- How runs work10 questions
- Security and data6 questions
- Billing3 questions
Product path overview
Getting started
Where do I sign in to ProPR?
On your own instance, with GitHub. Self-hosting is free, and ProPR Hosted runs a dedicated instance for you. The live demo lets you explore the product first.
Where are the API, docs and source?
Your instance serves the HTTP API under /api. Read the API reference and docs, inspect the GitHub repository, or ask on Discord.
What does ProPR need on the host?
A Linux Docker host. ProPR runs each agent in its own container next to the app and workers. Self-hosting covers the shape, and the deployment docs list the requirements. The guide to running AI coding agents on your own server covers what stays on your infrastructure.
Does ProPR work with GitLab or Bitbucket?
No. ProPR runs on GitHub pull requests, review comments, status checks, and comment commands.
Does it run against GitHub Enterprise Server?
No. ProPR runs against GitHub.com.
Do we have to use the hosted ProPR GitHub App?
No. The shared App through ProPR Connect is the fastest start. Bring your own GitHub App and you skip the relay entirely. Either way, ProPR asks only for contents, issues, and pull requests.
Can we run it from a phone?
Yes. Agents run on your server, so a phone only steers. Label an issue, review the diff, and leave a plain comment in the GitHub mobile app, and the follow-up commit lands on the same PR. See mobile and remote control.
Can we automate it with the CLI?
Yes. The CLI covers plans, tasks, goals, automations, repositories, agents, and logs, so scripts and CI jobs go through the same review path.
How runs work
Why keep the loop in GitHub?
Because that is where your team already reviews code. Agent output lands in the same pull request loop as human-written code: diff, comments, checks, approvals, merge.
Can ProPR work straight from GitHub issues?
Yes. Add a trigger label such as AI to an issue and ProPR implements it in an isolated run and opens a pull request linked back to the issue. A model label such as llm-claude-opus55 picks the agent, and several model labels open one PR per model to compare. Read the full issue-to-pull-request walkthrough.
What about changes too big for one PR?
Planning splits them into review-sized tasks, and Epic mode ships them as an ordered chain of pull requests, each merging before the next starts. The epic playbook walks through one.
Can we switch agents during the same change?
Yes. One agent opens the PR, another reviews it, a third takes the follow-up. The branch, task page, and pull request give every agent the same starting point.
How does ProPR decide what code to send the model?
It shows you before it sends anything. ProPR ranks files by git history, path signals, and short summaries, adds the files and attachments you name, and shows the selection and its estimated cost for you to adjust, so no run balloons into an enormous prompt. Context gathering.
Does indexing ship our code to a cloud index?
No. Indexing and file summaries run on your host with the agent you already configured. Repository knowledge stays in your stack.
Does ProPR read our CI logs?
Yes, through what GitHub reports for each check. With automatic CI follow-up on, a failing check starts a fix from the check's output, and Ultrafix waits for green before it continues. Or paste the failure into a PR comment and the fix lands as a follow-up commit. See how ProPR fixes failing CI runs in the same pull request.
Who verifies a resolved merge conflict?
Your CI does. ProPR resolves the conflict with a real merge commit and pushes it, and your checks run on it like any other commit.
What can we monitor afterwards?
Everything. Prompts, task history, commits, logs, model choices, cost, and follow-up activity stay on the task in the web UI and CLI, and GitHub keeps the branches and PR conversations.
Does ProPR clean up after itself?
Yes. Each run's Git worktree and Docker container go when the task completes. Task records and logs stay, so the audit trail survives until you prune it.
Security and data
How is each agent run sandboxed?
Every run gets its own Git worktree, branch, and Docker workspace, and the agent holds a read-only GitHub token. ProPR makes every commit, push, and pull request itself. Trust and security lays out the model in full.
Are agents cut off from the network by default?
No. By default, agent containers have normal outbound access. Optional restricted mode limits each run to an allowlist of hosts and lists every denied host on the task timeline. Trust and security covers where it applies.
Does connecting to GitHub open an inbound port into our network?
No. Your stack opens an outbound connection to ProPR Connect and GitHub events flow back down it, so a laptop or internal network needs no public ingress. Prefer direct webhooks? Run your own GitHub App.
Who holds the keys, and can we revoke access?
You do. With the shared App, your stack holds a relay credential you can revoke at any time, and the App's private key never touches your deployment. With your own App, the key is yours. Trust and security maps the full boundary.
Does ProPR Connect see our code?
Connect relays full GitHub webhook payloads, including any code quoted in an issue or review comment, clears each payload after successful delivery, and never runs agents. Trust and security covers retention and the hosted MCP route.
What stays on our infrastructure?
Repository clones, task history, logs, provider credentials, and agent containers. GitHub receives branches, pull requests, and comments; your chosen provider receives the task's prompt and code context. Trust and security shows who gets what.
Billing
What does ProPR cost?
Self-hosting is free and Apache 2.0. ProPR Hosted runs a dedicated instance on a Solo, Team, or Scale plan, and Connect seats are free for three with paid seats beyond that. Pricing has the numbers.
Do we have to use one provider or pay for new billing?
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.
Where do we see what a task cost?
In the usage log, which ties each call's model, tokens, cost, and duration to the task that used it. Cost and usage shows the log, and optional Agent Tank adds subscription capacity alongside it.
Run it yourself in two commands.
A running stack answers every remaining question, and the source is open.