Heavy Runs: done-for-you AI builds, audits, and research.
TL;DR: A multi-agent fleet takes the whole job and hands back the finished artifact: a merged-ready PR, a prioritized audit with fix PRs, or a cited PRD. Every run is expert-accepted before delivery. You bring your own LLM key. Founding pricing is open to the waitlist.
What happens in a BUILD / SHIP heavy run?
You hand off: a repo, branch rules, task brief, acceptance criteria, and any constraints the fleet must respect.
You get back: a working pull request with implementation notes, tests where fit, and the review trail needed to decide whether to merge.
Expert-accepted means: a human engineer checks the PR against the brief, failure modes, and repo conventions before it is delivered.
What happens in an AUDIT / FIX heavy run?
You hand off: a codebase, risk areas, production concerns, and the kind of fixes you are willing to accept.
You get back: a prioritized audit covering bugs, security, performance, and maintainability, with fix PRs when the selected lane calls for them.
Expert-accepted means: the findings are checked for reproducibility, impact, and false positives before they reach your queue.
What happens in a RESEARCH / PRD heavy run?
You hand off: a decision, market question, product direction, or technical unknown that needs a sourced answer.
You get back: a cited PRD, technical spec, or strategy doc with assumptions, open questions, risks, and source links.
Expert-accepted means: the document is checked for unsupported claims, weak reasoning, missing tradeoffs, and whether it can drive execution.
Questions operators ask before joining
What is a heavy run?
A done-for-you build, audit, or research job handled by an autonomous multi-agent fleet and expert-accepted before delivery.
What can come back from a heavy run?
A merged-ready PR, a prioritized audit with fix PRs, or a cited PRD or technical spec, depending on the family.
Do I bring my own LLM key?
Yes. The fleet runs on your provider key, so model usage stays on the account you control.
How is pricing handled?
Founding pricing is open to the waitlist. Public dollar figures are not posted for heavy runs.
Pitstop — one job, one run, one receipt.
Use the catalog when a fixed-scope run is enough.
AI Jungle OS — the private cockpit when the same job comes back every week.
Use the OS when the run needs to repeat with memory, schedules, and approvals.
AI Jungle, the agency — when it should be built and run for you.
Use the agency lane for custom buildout, audits, and managed operation.