MCP from chat
Plan, direct agents, inspect progress, and control the PR workflow from chat through direct instance access or Connect. Check MCP setup and availability.
Both routes use the same core tools. Agents and execution remain self-hosted.
Plan, run, review, and refine AI-generated pull requests.
ProPR is built around the moments that make AI coding useful in a real codebase: deciding what should change, giving the agent the right context, keeping execution isolated, opening the PR, and turning review feedback into the next commit.
Gather context
Execute
Refine the PR
Feature overview
The pieces that make the PR loop work.
Planning
Start with a repository-aware plan: indexing, context preview, refinement, empty-repo starts, and approval in one flow.
Planning turns a prompt into a reviewable scope before the agent runs.
PR fine-tuning
Keep iteration in GitHub. A normal comment asks for a change, and ProPR returns a follow-up commit on the same branch.
Agent iteration belongs where the team already reviews code.
Ultrafix
Ultrafix runs an automated review-fix loop on a PR until it hits the target review score, maximum cycle count, or a human stop signal.
Use this when a branch needs several rounds of cleanup pressure in a row.
GitHub comment control
Comments in the PR conversation request fixes, reviews, branch updates, conflict help, model changes, and recovery.
Once the first PR is open, its comment thread drives the next action.
Worktree isolation
Each task gets its own Docker environment, branch, and Git worktree, so the agent cannot mutate the main checkout directly.
Isolation is the concrete safety boundary, so agent reliability never has to be taken on faith.
Agent routing
Every supported agent runs through subscriptions or direct APIs on the same planning and PR path.
Compare this with direct agent use when the task is still exploratory.
Context gathering
Give the model selected files, summaries, screenshots, attachments, and reference material. Keep the context focused.
The context the model will receive is visible before any spend starts.
The product around the loop.
These features make ProPR usable across repositories, people, agents, and operating habits. They reinforce the PR loop.
Repository intelligence
Improvement suggestions and repository todos turn codebase knowledge into planned changes.
SummariesTodosPlan drafts
Operations and automation
CLI control, task history, logs, dashboards, and settings make the process scriptable and auditable.
CLILogsDashboards
Control and cost
Agent Tank, trust and security, and self-hosted access keep credentials, usage, and infrastructure choices visible.
Self-hostedUsageCredentials
Start with the product path.
New to ProPR? Start with planning, then PR fine-tuning in GitHub — the loop the rest of the product supports.
Go deeper when you have a specific concern.
Existing repositories explains ProPR's focus on repo-owned software. Engineering leads covers governance and rollout. Recipes shows concrete paths through the product.