Skip to content
Rush Commerce
Software & Dev3 min read

Replit Missions: agent swarms on a budget you set

Replit Missions runs AI agent swarms against a contract: success criteria, a USD budget, approval gates. The budget can run over. Write criteria you can check.

You don't prompt Replit Missions. You sign a contract with it. Replit launched Missions on November 17: you name an outcome, set a dollar budget and a list of actions that need your approval, and a swarm of AI agents plans, builds, tests and iterates until an AI judge says the goal is met or the money runs out. The pitch to small teams is direct. The project you never had headcount for now has a price tag.

What actually happened

Per Replit's launch post and Missions docs:

  • The mission contract. Before work starts you agree on four things: the approach, success criteria, a budget (maximum Replit AI usage in USD) and boundaries (actions that need approval, and actions it must never take). You also pick the model.
  • Approval gates. Replit's own example caps a sales mission at $1,000 and requires sign-off before agents contact prospects, share a proposal or make product commitments.
  • The judge. An automated evaluator checks the evidence against your criteria. If the budget runs out first, work stops with the goal unresolved.
  • The showcase. Replit says a swarm built DataCat, an in-house application monitoring platform, in two days on a $3,000 budget. It caught a planted database bug in 38 seconds against a 60-second goal and passed 51 of 51 data and alert checks. Replit is open-sourcing a self-hostable version.
  • Price. The post doesn't list a per-mission price or say which plans include Missions. It points teams to sales.

Every benchmark here is Replit's. Its own footnote says the comparison with a Claude Code build used one run per setup.

Why it matters for your business

The fine print is the useful part. The docs say work already running "can take AI spending slightly over the budget before it stops." They say automated completion "is not a guarantee that the result is correct." And clearing a mission doesn't cancel background processes or reverse actions already taken. A budget cap is a target with a soft edge, not a hard wall. Plan for that.

Your success criteria do the real work. The judge can only check what you wrote down. "Build a monitoring dashboard" can't be checked. "Alert within 60 seconds of a failed database query, zero lost log lines over a three-hour load test" can. Replit's docs make the same point: write criteria you can check, not a list of activities.

Build-vs-buy just moved. A $3,000 internal tool against a monitoring bill that renews every year is a real calculation. But the build is the cheap part. Someone owns patches, uptime and the next schema change. Price the ownership before you price the mission.

Key takeaways

  • Replit Missions launched November 17: agent swarms working to a contract you approve
  • The contract sets success criteria, a USD budget, approval gates and banned actions
  • The budget can run slightly over, and the AI judge's "complete" is not a guarantee
  • DataCat ($3,000, two days) is Replit's own test, one run per setup
  • No per-mission price published; Missions goes through Replit's sales team

Thinking about replacing a SaaS bill with something you own? Run the numbers first: build cost, plus the years of upkeep the agent doesn't do. Try our ROI calculator, or talk to us about writing the success criteria and owning the result.

Sources: Replit: Introducing Replit Missions, Replit Docs: Missions.

  • #replit
  • #ai-agents
  • #agent-swarms
  • #budgets
  • #build-vs-buy
TR

Tommy Rush — Founder, Rush Commerce

Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More

Get The Rush Report weekly — one email, zero fluff.