evals on your own work
pass96%
8/13 07:03fail0.4$0.39
8/13 07:02pass0.8$0.53
8/13 07:02pass0.8$0.46
$ introducing warp factories
Open infrastructure for cloud software factories
build on Warp: factories as code, any model or harness, with evals, benchmarks, and self-improvement built in.
get up to $10,000 in free factory usage
livefactory.yaml
# factory.yaml schemaVersion: v1alpha1 name: acme-web repositories: - owner: acme name: web mcpServers: github: warpId: mcp_github agentDefaults: model: claude-5-fable-high
trusted by 800k+ devs at
close the loop
Benchmark. Optimize. Repeat.
benchmark on your own work
Replay historical tasks from your factories to measure performance against your team’s real workflows.
measure what performs best
Compare models, harnesses, and configurations across quality, correctness, efficiency, and custom scorers.
put the results to work
Update your factory code to route work to the best-performing setup.
early access
Request early access
Set up your first factory with early access.
get up to $10,000 in free factory usage
SDLC coverage
Beyond CI/CD to automating the whole SDLC
defined in code, built for scale, and easy to deploy.
# factory.yaml — factories as code schemaVersion: v1alpha1 name: pr-review repositories: - owner: acme name: web agentDefaults: model: claude-5-fable-high # agents/foreman/agent.md agentType: FOREMAN # agents/review/agent.md agentType: REVIEW model: glm-5.2-fireworks # automations/on-pr/automation.md agent: review triggers: - provider: github event: pull_request_ready
# cli $ warp agent run-cloud --environment ENV_ID \ --prompt "fix LIN-482, low risk only" # api $ curl -X POST https://app.warp.dev/api/v1/agent/runs \ -H "Authorization: Bearer $WARP_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "prompt": "fix LIN-482" }' { "run_id": "run_4026", "state": "QUEUED" } # sdk (typescript) const run = await client.agent.run({ prompt: 'fix LIN-482' }); console.log(run.run_id); # mcp · warp-factory server { "tool": "send_task", "factory_uid": "fct_acme_web", "title": "Fix LIN-482", "note": "low risk only" }
# factory.yaml — factories as code schemaVersion: v1alpha1 name: pr-review repositories: - owner: acme name: web agentDefaults: model: claude-5-fable-high # agents/foreman/agent.md agentType: FOREMAN # agents/review/agent.md agentType: REVIEW model: glm-5.2-fireworks # automations/on-pr/automation.md agent: review triggers: - provider: github event: pull_request_ready
# cli $ warp agent run-cloud --environment ENV_ID \ --prompt "fix LIN-482, low risk only" # api $ curl -X POST https://app.warp.dev/api/v1/agent/runs \ -H "Authorization: Bearer $WARP_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "prompt": "fix LIN-482" }' { "run_id": "run_4026", "state": "QUEUED" } # sdk (typescript) const run = await client.agent.run({ prompt: 'fix LIN-482' }); console.log(run.run_id); # mcp · warp-factory server { "tool": "send_task", "factory_uid": "fct_acme_web", "title": "Fix LIN-482", "note": "low risk only" }
issue flow
How work moves through a factory
issues land from the channels your team already uses, the foreman moves them down the line, and the line stops wherever you decide.
a new idea or bug landsslack · linear · jira · github · API / MCP · schedules
foreman · waiting on humanneeds spec approvalyou review the spec
foreman · waiting on humanneeds code reviewyou review the code
foreman · waiting on humanneeds QAyou review the product
01 triageoptionalcause, scope, a tracked issue — skipped when the ask is clear
02 specoptionalthe plan for how the change should land — skipped for small, clear work
03 implementwrites the code and tests, opens a pr
04 verifyadversarial review, evals, and checks
05 shipthe approved change lands
06 monitorwatches prod, files what it finds
a new idea or bug landsslack · linear · jira · github · API / MCP · schedules
f the foreman moves each task stage to stage — and stays the one voice that reports back to you.
01 triageoptionalcause, scope, a tracked issue — skipped when the ask is clear
02 specoptionalthe plan for how the change should land — skipped for small, clear work
you review the specneeds spec approval
03 implementwrites the code and tests, opens a pr
you review the codeneeds code review
04 verifyadversarial review, evals, and checks
you review the productneeds QA
05 shipthe approved change lands
06 monitorwatches prod, files what it finds
⟲ monitor feeds what it finds back onto the line
the foreman (f) moves each task down the line · it stops where you decide · merge is always yours
case study
Cutting our cost per PR from $80 to $30
How we used Warp Factories Benchmarks to test models on our own engineering tasks and optimize for cost without sacrificing quality.
read case study
no lock-in
Open at every layer
layer 01
Any agent
warpclaude codecodexcursor
any MCP-capable coding agent.
layer 02
Any model
frontieropen-weight
frontier or open-weight, chosen per pipeline stage.
layer 03
Your compute or ours
warp cloudself-hosted
warp's cloud, or self-hosted in your own VPC.
layer 04
Data lives where you want
pluggablezero-retentionyour VPC
you own and store what your factory produces.
quality loop
Baked-in measurement and self-improvement
evals, benchmarks, and self-improvement loops drive measurable gains.
benchmarks across models
best1.00x
Claude Fable 5pass1.00x$0.53
GPT-5.6 Solpass0.94x$0.41
Gemini 3.6 Flashfail0.89x$0.32
self-improvement loops
auto-fix+3.7
memory updatedpass+0.2#4021
prompts tunedpass+3.2%#4022
regression caughtfail-0.4#4023
"warp factories drove our cost per agent pr down by 30%."
— vp engineering, series c infrastructure company
governance
Control your coding agent chaos
continuous improvement, better governance and security, by default.
increase agent roi · cost per pr
$—
computeplatforminference
one control plane for every agent
resolve merge conflicts on server pr 14090fable 5orchestrationfnow
refactor auth middleware for session reuseopus 4.8claude codez14m
investigate child process kill issuefable 5slackj37m
investigate usize panic in offset codegpt-5.6 solslackf1h
fix desktop tab crashgemini 3.6slackj2h
add rate limiting to public api endpointsgpt-5.3codexe4h
agents prove their work with computer use
every factory agent captures a screenshot or video so you can verify its work before shipping a PR.
Activations dropped on the contact sales page this week. Can you investigate and fix the culprit?
On it — opening the browser to test the flow myself...
Bring Warp
to your team
to your team
Enter your work email and answer one question — we'll get you to the right person on our team.
Book a demoWork Email*you@company.comCONTINUE
Trusted by over 800,000 developers and thousands of engineering teams at leading companies
Found it — the testimonials section was hidden by collapsing divs. Patched, PR #436 is up for review.
built to scale
The factory infrastructure you'd build yourself
available now, so your eng time goes into your product, not the plumbing. tick a box to clear the desk (drag the pile around on desktop).
TO_DO
Slack and Teams event webhooks
TO_DO
Jira and Linear issue webhooks
TO_DO
GitHub and GitLab push/PR webhooks
TO_DO
webhook signature verification
TO_DO
cron scheduler for triggered runs
TO_DO
event-to-task deduplication
TO_DO
per-run sandbox provisioning
TO_DO
Docker- or Kubernetes-backed compute
TO_DO
environment image build and caching
TO_DO
cold-start latency for new runs
TO_DO
sandbox teardown and cleanup
TO_DO
multi-harness process management
TO_DO
Claude Code, Codex, Cursor, and Warp support
TO_DO
model routing per pipeline stage
TO_DO
MCP server connection management
TO_DO
CLI invocation inside the sandbox
TO_DO
source-forge integration: GitHub, GitLab
TO_DO
non-human identity per agent
TO_DO
credential and token handling
TO_DO
fine-grained permissions per agent
TO_DO
secrets storage and rotation
TO_DO
quotas and concurrency limits
TO_DO
network egress policy enforcement
TO_DO
run lifecycle state machine
TO_DO
durable execution across long-running steps
TO_DO
retries and idempotency
TO_DO
multi-agent handoff between stages
TO_DO
workflow versioning
TO_DO
thread and run continuity
TO_DO
real-time log and event streaming
TO_DO
run history and audit trail storage
TO_DO
audit trail per tool call
TO_DO
live session attach for steering
TO_DO
failure detection and alert routing
TO_DO
approval-gate enforcement before merge
TO_DO
checkpoint pause/resume state
TO_DO
evals and scoring pipeline
TO_DO
artifact and output storage
TO_DO
benchmark regression detection
TO_DO
…and everything you haven’t hit yet
Frequently asked questions
a factory is a fleet of agents wired to your SDLC — triggered by an issue, a slack message, or a schedule, and coordinated by warp factories from triage through review to a mergeable PR.
as code, in a factory.yaml + supporting agent files. define triggers, agents, models, and approval gates — check it into your repo like any other config.
no. bring your own model or harness (e.g. claude code or codex) — warp factories works with whatever your team prefers and helps identify the most cost-effective configuration over time.
no. factories is a separate product — nobody at your company has to use warp terminal. work comes in through tools you already use, like slack, linear, jira, github, and gitlab. if you do use warp terminal, it has built-in factory integrations (native mcp) so you can iterate locally.
most orgs start around 20-30% of PRs fully automated, starting with simple tasks. over time this goes up as models improve and your factory self-improves.
wherever you choose — warp's cloud, or fully self-hosted inside your own VPC, under your existing retention and compliance rules.
humans put work into the factory via your existing tools like slack, teams, linear, jira, github and gitlab. at any time you can iterate on that work locally and the factory agents loop you in proactively when they need help.
yes — most teams start with a single low-risk workflow, like dependency bumps or flaky test triage, before expanding coverage.
usage-based, priced per agent run. qualifying orgs get $10k of factory usage during closed early access.