Artificial intelligence news.
Straight from the original announcements.

Replit launches Missions to coordinate AI agents on complex business goals

November 17, 2026, 18:31By the AI News Bites Editorial TeamSource: Replit
Replit launches Missions to coordinate AI agents on complex business goals
Image: Replit

Replit has launched Missions, a service that coordinates swarms of agents to pursue complex business goals, from winning enterprise customers to replacing legacy software. Users define the outcome, budget and boundaries, while the agents plan, build, test and iterate toward it.

Key points
  • Replit Missions coordinates agent swarms around user-defined goals.
  • Users control budgets, boundaries and approval requirements.
  • Replit used Missions to build the DataCat monitoring platform in two days.
  • Replit says DataCat passed 51 of 51 data-correctness and alert-reliability checks.
  • A self-hostable version of DataCat is being open-sourced.

Users set the goal, budget and permissions

Missions is designed for work that teams do not have the time or capacity to handle. A user could define a mission such as winning the first 10 paying enterprise customers without deciding in advance which industries to target, which buyers to contact or which customer problem to prioritize. During setup, Missions asks how success should be measured, including whether paid pilots count or whether the goal requires signed annual contracts above a minimum value.

The service also uses a Missions agreement to establish a spending limit and define actions that require approval. Replit gives the example of a $1,000 cap that requires permission before agents contact prospects, share a proposal, change customer-facing sales materials or make product commitments. Those restrictions remain in place as the mission evolves.

Agent swarms work in parallel and adapt

Agents can research accounts, build tailored demonstrations and prepare outreach for review. They can work in parallel, then use responses and shared findings to guide subsequent work. For example, a mission might focus on an industry showing stronger interest, revise a demonstration around a recurring buyer question or change messaging that receives few responses.

Replit says agent swarms can also help with optimization by exploring several implementations at once. In 13 NumPy benchmark tasks, a single Auto (Max) agent achieved speedups of up to 4×, compared with 17× for a 16-agent swarm and 19× for a 32-agent swarm. With Auto (Power), the corresponding figures were 2×, 7× and 9%. Replit says the tests showed how parallel exploration could help find faster implementations, while also validating correctness.

Users can choose whether to receive blockers, progress updates and general status immediately, in a daily digest or weekly. An independent AI judge checks results against the agreed success criteria, and the mission is marked complete when those criteria are met. If the budget runs out first, the work stops unresolved.

Replit used Missions to build DataCat

Replit says it used Missions to build DataCat, an application monitoring platform intended to run on and be customized for its own infrastructure. In two days and within a $3,000 budget, the agents built dashboards, alerts, application tracing, an agent for collecting monitoring data and tracing libraries. Replit says DataCat raised an alert for a deliberately introduced database bug in 38 seconds, beating a 60-second target, and traced the failure to the exact database query and its logs.

During a three-hour load test, DataCat stored 26 million application activity records and 4.4 million log entries with no data lost. Replit says all 36 data-correctness checks and 15 alert-reliability checks passed. The company also reports 51 of 51 checks passed in a broader summary, 39% more requests handled with 51% less monitoring-related processor overhead in matched Node.js tests, and 49,000 traces per second with DataCat using Postgres, compared with 9,800 for a Claude Code build.

Replit cautions that the controlled tests do not establish performance against the original platform’s hosted service or on Replit’s production systems. The company says it is open-sourcing a self-hostable version of DataCat, giving users a concrete example of the kind of work Missions is intended to support.

ReplitReplit MissionsDataCatReplit SlidesNumPyAI agentsenterprise AIcoding

Related news

Databricks Opens Genie ZeroOps to Public Beta for Pipeline Fixes
Products and toolsNovember 19, 2026, 22:30

Databricks Opens Genie ZeroOps to Public Beta for Pipeline Fixes

Databricks has moved Genie ZeroOps into public beta, giving workspace teams an autonomous agent that detects, investigates and helps resolve data pipeline issues. The…

Baseten says deferred tool loading cuts agent costs by 62.5%
For developersNovember 19, 2026, 18:49

Baseten says deferred tool loading cuts agent costs by 62.5%

Baseten says its deferred tool loading support for open models reduced coding benchmark costs by 62.5% and cut task times by 26.2%, without changing the model or…

Replit launches creator program for B2B voices
BusinessNovember 18, 2026, 19:56

Replit launches creator program for B2B voices

Replit has launched Replit & Co., a creator program designed to bring B2B voices into what the company calls the era of the self-driving company. The announcement was…