ProPR
Free, Apache 2.0, self-hosted. You pay providers directly — subscriptions or API keys — and an optional hosted webhook relay (ProPR Connect) adds GitHub convenience.
Free softwareNo token markupUsage tied to tasks
AI coding bills come in four shapes. Learn to read each one.
AI development tools rarely cost only what the pricing headline says. The real bill depends on seats, model usage, premium requests, cloud tasks, background agents, code review volume, infrastructure, and the engineering time spent turning AI-generated changes into pull requests ready for review.
vs
AI coding bills
ProPR cost model
The software is free and open source (Apache 2.0). You run it on your infrastructure and pay model providers directly — through subscriptions you already hold or your own API keys. ProPR never marks up tokens.
Current tool surfaceEditor, hosted workspace, review tool, or trackerWork starts where that product keeps its strongest context.
ProPR pathPlan → isolated workspace → GitHub PR → follow-up commitThe pull request and its review conversation preserve the shared history.
Exact prices change quickly. Billing shape changes more slowly. A low-cost editor seat can become expensive when background agents consume extra credits. A hosted coding agent can look simple until heavy tasks hit rate limits. An open-source agent can be free to install but expensive to operate if every repository, sandbox, secret, retry, and PR update has to be wired in-house.
ProPR's model is deliberately simple. The software is free, open source (Apache 2.0), and self-hosted — it owns plans, tasks, isolation, pull requests, comments, logs, and recovery. The agent bill stays yours and stays direct: use the AI provider subscriptions you already hold (Claude Pro/Max, ChatGPT Plus) or your own API keys, with no token markup in between. ProPR Connect is an optional hosted webhook relay for GitHub event routing and replays, free for a few seats and paid beyond.
Start with what the bill is buying.
Four shapes cover most of the market. Seats price people, whatever they use. Credits price provider-hosted runs in the vendor's own units. API tokens price raw model usage directly. Subscriptions price flat-rate access to one provider's agent. Most tools bundle several of these; ProPR unbundles them.
Hosted agents and AI IDEs
Usually seats, credits, premium requests, background-agent usage, cloud tasks, or provider-specific limits.
Fast adoptionProvider workspaceLimits vary
Direct agents and APIs
Subscriptions or token billing, plus the engineering time needed to turn private sessions into branches and PRs.
FlexibleManual handoffToken-sensitive
A cheap line item can hide an expensive path.
| Cost source | Usually hidden in | How ProPR frames it |
|---|---|---|
| Agent/model spend | Credits, premium requests, token bills, or subscriptions | Chosen per task |
| Cloud execution | Hosted agent runs and background tasks | Your infrastructure |
| PR handoff | Manual cleanup, PR writing, and review follow-up | Built in |
| Open-source maintenance | Queues, sandboxes, logs, secrets, and retries | Productized |
The hidden cost is usually handoff.
Private sessions still need packaging
Direct agents and IDEs can produce useful code, but someone still has to clean the branch, write the PR, explain the change, and handle review comments.
Review-only tools stop after the comment
They can reduce review effort, but the fix often moves back to a developer or another agent session. ProPR keeps PR feedback on the same branch.
Open-source stacks move cost into maintenance
No license cost can still mean queues, sandboxes, secrets, logs, retries, PR automation, and support owned in-house.
Use public prices as anchors and confirm exact numbers with each vendor.
Vendor prices change often, so check current numbers on their pricing pages. Compare the billing shape: seats, credits, API usage, cloud runs, PR volume, infrastructure, and the time saved when follow-up fixes remain attached to the same PR. ProPR's own numbers live on the pricing page; model spend goes directly to the provider.
SeatsCreditsPremium requestsAPI tokensCloud tasksReview volumeSelf-hosting
Low sticker prices can hide high total costs.
Agentic coding can burn cost beyond seats: long contexts, repeated review cycles, branch cleanup, manual PR writing, failed attempts, and hidden cloud task limits.
- Ask how a long-running task is billed.
- Ask whether retries and follow-up changes consume the same quota.
- Ask who pays for compute, storage, and logs.
ProPR makes the cost boundary explicit.
The process lives in your deployment. The agent/model bill is chosen per phase and can change as pricing, model quality, or subscription access changes.
- Use subscription-backed agents where they are economical.
- Use API billing where direct usage control matters.
- Switch models without changing the PR path.