Introducing Warp Factories: open, flexible infrastructure for building cloud software factories. - Configure your factory as code - Use any model and any harness - Measure quality with evals and benchmarks on your own data - Built-in self-improvement and memory

Factories are defined as code and built to integrate with the tools you already use. Think Terraform for agent configuration.

Work enters your factories from the tools your team already uses: - Communication tools like Slack or Teams - Task trackers like Linear or Jira - Source code forges like Github or Gitlab - Terminals, IDEs and other local coding agents via the Warp Factory MCP

Bring whatever model or harness is best for your workflow. You can use Warp’s SOTA agent for multi-model access, including open-weight models, or you can directly run Claude Code or Codex as the harness.

All of these agents have access to computer use on Linux and Mac to reproduce issues and prove correctness of changes. The agent can share recordings on pull requests and within conversations, so you never need to pull down code to verify a change. Here's a real example:

Measure factory performance with cost and velocity metrics, accessible from the factory dashboard and via our API and SDK. Plus, configure self-improvement agents to grade performance and automatically improve your factory overtime.

We're onboarding a limited number of companies onto Warp Factories today. For select folks, we are offering $10k of factory usage to help get you started.​​​​ 👇

@grok explain to me like i’m a fucking idiot

whoever did the design deserves a raise ngl

factories as open infra >>> factories as some black box product

Incredible product presentation. Oh and factories are cool too

@sohumgaitonde @varoonkodithala @vikvang1 LETS GOOO

Incredible

The evals-on-your-own-data part is the piece most agent setups skip. Config as code plus benchmarks is what makes it reviewable instead of vibes — that's the difference between an experiment and something I'd let near a production repo.

Any model plus any harness plus evals on your own data is the real unlock, and also the trap. If both vary at once a regression cannot tell you whether the model drifted or the harness changed. Pin the harness as a control. #LLMOps @warpdotdev

Can I bring my own cloud VM??? @asciidotdev

config as code beats vibes per repo

this is next level

This is the upgrade teams actually need. Factories as code. Any model. Real evals on your data. Self-improvement built in. And it plugs into the tools we already use instead of forcing a whole new stack. Finally, something that adds real value to the workflow. Feels right.

Flexible infrastructure like this could reshape how software teams scale

@grok is there open source alternative for this ?

the evals on your own data are the actual product. the agents were never the bottleneck.

Warp Factories turn infra into code that improves itself

looks nice but what part of this is open? it's closed source and waitlisted sign ups

Any model plus any harness, written as code. The factory is the product. The chat window was the demo.

what is a software factory. what is a software factory thats builds software factories? is this like that one time peter steinberg said you should run all your software in loops.

Cloud software factories, cool concept 🏭

so is warp a terminal, agent platform or something else 🥺

Treating infrastructure 'as code' with built-in self-improvement is the right direction. Most teams still manually tune their AI workflows, having the system evaluate and refine itself against your own benchmarks turns this from a one-time setup into a continuously improving system.

where can we report bugs? why Warp drops the Shift modifier on the numpad Enter key specifically?

@grok can I use my existing OpenAI/ Claud subscription or I am forced to use theirs ?

This is a compelling direction. AI software factories with real flexibility, evals, and memory.

The useful shift is making the factory itself inspectable. Config as code plus evals on your own data turns self-improvement from a slogan into a loop a team can measure and change.

Looks awesome, can I bring my Grok subscription?

The @FactoryAI is the product 👀

Will it be open source ?

Brilliant concept factories as code with self improvement is the future

What does the word open mean here?

evals on your own data is the hard bit. how do you stop them rotting?

evals on your own data is when a coding agent becomes a factory

SoftwareFactoryFactoryImpl