Warp launched Warp Factories on Tuesday, a turnkey infrastructure layer for building and running the kind of AI-driven software factory that companies like Stripe and Ramp have spent the past year assembling by hand. The system slots into the standard phases of software development — triage, specification, implementation, review, and verification — and lets teams automate any of those stages with agents running in a shared, observable environment. Warp CEO Zack Lloyd says his own team already automates 30% to 35% of its weekly tasks using the approach.
The pitch is aimed squarely at the middle of the market. Stripe has publicly documented its internal "minions" system for automating work inside its codebase, and Ramp has built a background agent that monitors code after deployment. Both took serious infrastructure investment to stand up. Factories is what you buy when you don't have that engineering budget but still want the same shape of workflow.
Users pick their own coding model and harness. The system is designed to work equally well with Codex or Claude Code, and it hooks into ticketing tools like Linear and Jira as well as messaging in Slack and Teams. The idea is to plug the agent loop into whatever workflow already exists, rather than force a rewrite.
Key facts
- 01Warp Factories launched Tuesday as an infrastructure layer for building and running AI coding agent loops.
- 02The system supports both Codex and Claude Code, plus Linear, Jira, Slack, and Teams integrations.
- 03Warp CEO Zack Lloyd says the company automates 30% to 35% of its own tasks on a weekly basis.
- 04Target customers are smaller companies without the engineering headcount to build a Stripe- or Ramp-style factory in-house.
That flexibility matters because the real work of running an agent fleet is not the model call — it's everything around it.
Beyond execution, Factories gives managers a dashboard for tracking how the factory is performing. Because every agent runs in the same environment, teams can compare configurations against each other, watch token spend, and identify which harness or model combination is producing the best results on which type of task. Warp also builds in self-improvement loops that let the system tune its own configuration over time.
This is the part that separates a real software factory from a chat interface with a coding model attached. Evals, memory that persists across agent runs, and cloud-side steering are the load-bearing pieces, and they're the pieces most teams underestimate when they first try to deploy coding agents at scale.
Lloyd is careful not to oversell the automation ceiling. Factories is built to let engineers collaborate with an agentic workforce, not replace it — and even at Warp, a company running its own product internally, roughly two-thirds of tasks still require a human in the loop.
Warp is betting that number moves. As frontier models get better at long-horizon coding tasks and as harnesses like Factories improve at supplying the right context, the automatable share of an engineering team's backlog should keep climbing. The company's product roadmap is essentially a wager on that curve.
The competitive frame is worth naming. Every major coding-agent player — from the model labs shipping first-party harnesses to the IDE vendors bolting agents onto existing editors — is racing to own the layer between the model and the codebase. Warp's angle is that the orchestration, evals, and telemetry layer is where the durable value lives, not the model or the editor. If a customer can swap between Codex and Claude Code without touching the surrounding infrastructure, the infrastructure is what they're paying for.
The open question is whether smaller companies actually want a prescriptive factory architecture or would rather assemble their own from cheaper primitives. Stripe and Ramp built bespoke systems because their scale justified it; a 40-engineer startup may or may not need the same shape. Warp is betting the answer is yes, and that the buyers who can't afford to hire a platform team to build agent infrastructure will pay for one that ships pre-wired.
For the broader AI coding market, Factories is a marker of where the frontier of the product category is moving. The interesting fight is no longer which model writes the cleanest function — it's which vendor owns the loop that dispatches, evaluates, and improves the agents doing the work. If Warp's 30-to-35% internal number is a preview of what mid-market engineering orgs can hit with off-the-shelf tooling, the buying decision for CTOs stops being about individual coding assistants and starts being about which factory floor to stand up.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
