Every token, on the record of the change that spent it.

ProPR is free and never marks up a token. You pay providers directly through the subscriptions you already hold or metered API billing, and ProPR shows the model, tokens, and cost behind every pull request.

What you pay for

Self-host with your own GitHub App and provider usage is the only cost beyond your infrastructure. Connect charges for Plus seats beyond three users, and Hosted is a paid plan; see pricing.

The subscription is only part of what AI development costs. Long contexts, retries, premium models, background agents, manual PR preparation, branch cleanup, review follow-ups, and rebuilding lost context all add up.

ProPR puts each of those costs on the record. Every task has a plan, an agent, an execution record, a pull request, logs, usage, and follow-up history. With Agent Tank on, ProPR ties subscription capacity to the change that consumed it.

LLM log rows tied to their work: two pull requests, a plan and a repository summary, each with its tokens, cost and the capacity its call used — Each model call names the task, plan, or repository it served, with its cost and the subscription capacity it used.

The bill has three parts.

ProPR (free)

Plans, tasks, GitHub integration, PR follow-up, logs, settings, CLI, and task history. No license fee and no seat charge when you self-host with your own GitHub App.

Your infrastructure

The Docker host, storage, database, repositories, and worker capacity you run.

Agent usage

The model spend behind each task, on your existing AI provider subscriptions or direct API billing.

Spend the stronger model where it pays off.

Put planning, careful implementation, inexpensive review, and PR follow-up on different providers. ProPR keeps one branch and one history across all of them.

Planning models per phase: Claude Opus 5.5 analyses plan context and Claude Fable 5.1 writes the plan — Per-phase models: Claude Opus 5.5 gathers plan context, Claude Fable 5.1 writes the plan.

Use subscriptions you already have

Claude Code, Codex, and Antigravity CLI run on the provider subscriptions you already hold; OpenCode and Mistral Vibe run on their own plans or keys. Provider plan limits apply.

Quota-awareGood for routine tasks

Use direct APIs when metering matters

Bring your own API keys for provider-level billing, model-specific control, and exact per-task accounting.

Clearer spendProvider controlsTask-level logs

Choose the model by phase

Use the model that fits the phase, then keep follow-up commits, retries, and review history on the same PR.

PlanningImplementationReviewFollow-up

What did this pull request cost?

ProPR ties model calls to tasks and repositories, so you see the implementation, review, and follow-up usage behind each pull request. The LLM log records model, tokens, cost and duration. Agent Tank tracks subscription capacity for Claude Code, Codex and Antigravity. The guide to what each AI-written pull request costs walks through those views.

The LLM log: each model call with its model, tokens and cost, from a $1.31 Claude Opus 5.5 implementation to a three-cent repository summary — Each model call with its model, tokens and cost.

Cap it, too. Set a spend limit per task, per repository in .propr/workflow.yml, or for the instance. At the cap, ProPR stops the run, publishes the partial work as a pull request, and tells you in the Inbox and on GitHub.

Task timeline: the run reaches its spend cap, an estimated $5.03 of a $5.00 cap from .propr/workflow.yml, and ProPR stops the agent and creates the pull request — The run reached its $5 cap from `.propr/workflow.yml`. ProPR stops the agent and creates the pull request from the partial work.

Compare against the full delivery cost.

A cheap seat gets expensive when your team spends hours turning private agent output into branches, PR summaries, follow-up prompts, and recovery steps.

Who pays for retries and long-running tasks?
Do review follow-ups land on the same branch?
Does usage tie back to the change that consumed it?

Compare ProPR in market context.

The cost comparison places ProPR beside hosted agents, AI IDEs, app builders, open-source orchestration, direct agents, and PR review tools.