Ship the feature.
See the lift.
Foredeck is the stats and control plane for your software factory. Your pipeline measures everything up to the merge. Foredeck measures what comes after: which run, prompt and model wrote the code, what the deploy did, and whether it survived production. Ramp it by percentage and turn it off without a deploy. Your agents keep running in your stack; Foredeck never runs them.
The loop
Report. Measure. Control. One object: the Change.
Every change becomes a Change
Report changes from CI through the Changes API or the SDK's factory module. Foredeck records the commits, gate verdicts and deploys behind each one, and marks it human, agent or mixed from commit trailers, bot authors and branch names. A GitHub Action is coming soon.
See what it did in production
Agent runs arrive as OpenTelemetry spans, with tokens and cost. Deploys, alerts and service metrics are stamped with the release, so each Change ends as survived, regressed or reverted. The factory scorecard, with change survival rate and cost per surviving change, is coming soon.
Guard the rollout
Foredeck Flags roll a change out by percentage and turn it off without a deploy, and alerts name the release that set them off. Coming soon: auto-rollback on a regressing guardrail, per-pipeline kill switches, spend caps, and outcomes sent back to your factory.
The platform
Works where your team — and your agents — work.
JavaScript and TypeScript, no lock-in
One SDK for Node, browsers and edge runtimes, with local Foredeck Flag evaluation and an OpenFeature provider for server and web, so you can adopt Foredeck without rewriting your integration.
An MCP server for your agents
From Claude Code, Cursor or your own agents: search logs and traces, query metrics, SLOs and UI events, list deployments, create and mute monitors, and acknowledge incidents. Foredeck Flag and Change tools are coming soon.
One pipeline for runs and services
Send agent-run spans over OTLP with the OpenTelemetry GenAI attributes, on the same trace pipeline as your services. No second pipeline, no second source of truth.
Runs on itself
We watch Foredeck with Foredeck.
Every service behind this page is instrumented to report through the same public ingest contracts you would use: metrics, logs, traces, and the monitors that page us. Our own factory is next: every PR to our repo will become a tracked Change, with the live numbers on this page.
That is also why measuring a change does not need a second vendor: the change, the release, the request latency, the error log, the trace behind it, and the alert are one system. Ship the change, and the evidence is already there.
What it's instrumented to watch
- Requestsrate · errors · p95, per endpoint
- Tracesbrowser → API → queue → consumer
- Logsstructured, correlated to the trace
- Queue laghow far behind the pipeline is running
- Alertscorrelated to the release that caused them
Dogfooding is not a slogan here: standing this up found two real bugs in our own ingest path.
Pricing
Start free. Grow when you're ready.
For a small team's month, or a factory pilot.
- About 100 tracked changes a month
- Changes, attribution and outcomes
- Unlimited Foredeck Flags, all environments
- Basic A/B readouts — lift + confidence intervals
- Capped monthly events
- Community support
For teams running pipelines and experiments continuously.
- More tracked changes a month
- Higher event volume
- Unlimited concurrent experiments
- Coming soon: sequential testing, guardrails + auto-rollback, spend caps, factory experiments
For factory scale and statistical depth.
- The largest tracked-change and event allotments
- Coming soon: CUPED / variance reduction, warehouse-native metrics, advanced segmentation, SSO + audit logs
Experiments are never behind the paywall. Two meters, never seats or MAU: monthly events, and tracked changes (a Change that merges in the month, whether a human or an agent wrote it). Past the allotment, new changes stop being tracked and Foredeck Flags keep serving. We're early, and pricing beyond the free tier is still taking shape with our design partners. Talk to us about what you need and we'll work out a fair price together.
For growth teams: how we compare
Foredeck Flags, with the experiment built in.
| Foredeck | LaunchDarkly | |
|---|---|---|
| Free, unlimited flags | Yes | Limited free tier |
| Experiments included free | Yes | Paid add-on |
| What's metered | Events and tracked changes; never seats or MAU | Usage-based |
| OpenFeature provider | Yes, server and web | Yes |
| Starting cost | Free to start | Free tier, then paid plans |
Why now
Every factory stops at the merge.
Software factories measure cost per PR, merge rate, tokens and eval scores. After the merge, the only signal most of them track is reverts. Foredeck covers the stations after the merge (the release, the rollout and the production outcome) and stays independent on purpose: a mid-market-native vendor built by founders who answer your emails.
Moving your flags off another platform?
We'll help you migrate your flags, experiments and metrics for free, with direct founder access during onboarding.
Talk to us about migratingGet in
Come build the factory loop with us.
A small number of design-partner spots, direct access to the people building this, and a real say in what ships next.