See spend as it happens
Each step shows what it used and what remains of the limit.
- Plan and scope$6.40
- Build checkout flow$18.25
- Run test suite$4.90
- Judge review$2.15
Missions
Subscription growth · $500 shared limit
Where paid conversion dropsSubscription growth · $500 shared limit
Pricing → signup → checkoutSubscription growth · $500 shared limit
Grow paid subscriptionsSubscription growth · $500 shared limit
One shared budget · $188 remaining
Subscription growth · $500 shared limit
Subscription growth · $500 shared limit
Evidence for the growth missionPaying customers
128KMonthly revenue
$256KPerformance
380msSubscription growth · $500 shared limit
Subscription growth · $500 shared limit
$312 of $500$188 remaining 62.4% used
Ask before increasing the $500 limit.
Subscription growth · $500 shared limit
Subscription growth · $500 shared limit
Approve subscription changes?Subscription growth · $500 shared limit
Subscription growth · $500 shared limit
Set a spending limit for each mission and track usage as agents work. Review any increase before it goes ahead.
Each step shows what it used and what remains of the limit.
Move different projects forward in parallel. See which missions are done, running, or waiting on you.
Review the reason for a higher budget, then decide whether to approve it or keep the current limit.
Set the outcome, review the plan,
and follow progress as agents get to work.
Benchmark and iterate toward an agreed response-time target.
Improve answers against a rubric and held-out test set.
Improve landing pages and lead capture against a conversion baseline.
Rewrite a component and verify parity with the original.
Recreate availability, reservations, rescheduling, and cancellations.
Audit and improve your site for search engines and AI answers, then measure visibility and qualified traffic.
Improve judging consistency against reviewed examples.
Find bottlenecks and measure improvements under consistent conditions.
Recreate drafting, approvals, signature handoff, and status tracking.
Pursue important work you haven’t had the time or capacity to tackle. Define success, then let Missions plan and coordinate the work.
Agents build, test, and iterate together. An independent AI judge checks results against your criteria, while progress updates help you see where things stand.
Parser batching
3 agents · example thread
Agent 1
Agent 2, test token batching. Agent 3, check buffer reuse. Keep partial Markdown rendering as it streams.
Agent 2
Batching preserves partial updates and the final syntax tree.
Agent 3
Both patches are isolated and ready for verification.
Eval judge
Agents do the work. A separate AI reviewer checks it against the goals you agreed on and sends back specific feedback when something needs fixing.
The proposal is missing its data retention terms.
In this example, the agreed goal requires the proposal to explain how long customer data is kept and when it is deleted. The reviewer finds that neither is covered and sends the draft back. The agents add the missing terms, and the reviewer checks the revised proposal before it reaches you.

Build a repeatable evaluation pilot to detect improvements and regressions in your customer-support assistant.
Success criteria and scope confirmed.
Identify the assistant and policy sources —
Assistant and policy sources have not yet been provided.
Awaiting source-grounded cases before scorer calibration.
Progress updates
As the mission runs, your agent writes short updates so you can follow progress and respond when your input is needed.
Type /mission, accept a suggestion from your agent, or start from the overview. The mission stays where the idea began.
Connect the tools you already use so the agent works inside your docs, tickets, and repos with the permissions you set.
Changes arrive as pull requests and reviewable files. Anything outside the agreed boundaries waits for a person.
Replit Missions uses agent swarms to turn ambitious goals into outcomes across complex builds, optimization, and business growth. It is designed for entrepreneurs, SMBs, and enterprise teams with work that takes multiple steps and iterations to complete.
A mission works toward a defined outcome, such as improving signup conversion. A routine runs on a schedule, such as sending a weekly signup report.
Describe the outcome you want to achieve. Work through the approach, success criteria, and boundaries, then review the mission agreement before execution. A useful starting goal names both the work and how you will judge the result, such as rebuilding a workflow against a feature checklist.
An agent swarm is a group of agents coordinating work toward a shared goal, with parallel work where appropriate. For a software build, that work can include implementation, testing, and evaluation, with results informing the next iteration.
Each mission has a judge that evaluates the work against the success criteria you agree on at the start. The mission keeps working until those criteria are met, its budget runs out, or it needs your input.
Set a maximum budget for your mission upfront. Work performed during the mission counts toward that budget.
You can set a maximum budget, decide which decisions require your input, and choose which connectors the mission can access. You agree on the goal and success criteria before work begins, and you can exit mission mode at any time. Messages you send while the mission is active count toward its budget.