Agent-native CI, with a bill you can read

Per-minute pricing for every job your CI and agents run, at the same rates on Latchkey CI and Latchkey on GH. You pay a plan fee, plus minutes and seats past what it includes.

Plan$5 to $49 /moeach includes free minutes
PlanPriceFree min / mo
Developer$52,000
Launch$194,000
Scale$496,000
EnterpriseCustomCustom
Minutesfrom $0.0025 /minonly once your free minutes are used
SizevCPURAM/ min
small28 GB$0.0025
medium416 GB$0.0050
large832 GB$0.0100
xlarge1664 GB$0.0200
Seats$5 /seatonly past included seats
PlanIncluded
Developer1
Launch3
Scale5

Developer is single-seat.

Free runner minutes are counted in small-equivalent units: a minute on latchkey-small counts as one, medium as two, large as four, and xlarge or a custom runner as eight, and the included allowance on every plan is unchanged.

latchkey-small 1×, medium 2×, large 4×, xlarge and custom 8×

Pick the plan that fits your team

Every trial starts on Launch, and no paid plan starts until you choose one.

  1. 1 · Start freeEvery trial starts on Launch30 days, no credit card. Sign up and you are running.
  2. 2 · Try it at full size10 repos, 3 seats, 4,000 free minutesEverything on this page is on, including the agent CLI and MCP server.
  3. 3 · Choose your planDeveloper, Launch or ScaleChoosing a plan starts billing right away and ends any time left in your trial. Developer holds one repo and one member.

Latchkey CI runs on personal accounts and organizations. Latchkey on GH, our managed runners for GitHub Actions, needs a GitHub organization.

CLI job context: up to 20 GiB of uploaded context per workspace in total at any one time, by default.

Launch

$19 /month
For early-stage teams with a few services.
Free minutes
4,000 / month
Repositories
Up to 10
Seats included
3
Extra seats
$5 / seat / month
Invite teammates
Yes
Jobs at once
40
Custom runners (Latchkey on GH)
2
Agent CLI + MCP
Included
CLI jobs per hour
240
Start free trial30-day free trial of Launch. No credit card.

Pricing calculator

Tell us how much CI you run and how many people need the dashboard. We'll pick the cheapest plan that fits.

How many repositories?
5
12,000
Share of minutes by runner size
smallthe restsmall 100%
medium
medium 0%
large
large 0%
xlarge
xlarge 0%
Dashboard seats
3
Your estimate
Launch
Launch plan
$19.00
monthly fee
Free minutes
$0.00
4,000 of 4,000 small-equivalent minutes
small
$20.00
8,000 min × $0.0025
Seats
$0.00
3 needed · 3 included
Estimated monthly total$39.00 /mo

Assumes all small at list rates, with free minutes spent in small-equivalent units on the cheapest size first.

Billed per minute, rounded up per job.

Start free trial →

30-day free trial of Launch. No credit card.

Moving from GitHub-hosted runners?Compare your bill with GitHub-hosted runners →

Per-minute rates, by size

Set runs-on: in your workflow file, or pass --size to the CLI. Each job gets a fresh VM with self-heal, at the same rates on Latchkey CI and Latchkey on GH, and a free minute counts more on a bigger size.

SizevCPURAMStorageLatchkey / minGitHub / minLess than GitHub
small28 GB100 GB$0.0025$0.00658%
medium416 GB100 GB$0.0050$0.01258%
large832 GB100 GB$0.0100$0.02255%
xlarge1664 GB200 GB$0.0200$0.04252%

On Latchkey on GH the runs-on: label is latchkey- followed by the size. Billed per minute, rounded up per job. Windows, macOS, arm64 and GPU runners are not offered today. See every provider side by side →

Rated A, and best value of ten platforms, by RunnerBench, which tracks published runner prices and dates every figure.

CI your agents can drive

Give an agent an API key and it can verify its own changes, pick up CI failures and run jobs. What an agent runs bills at the same per-minute rates.

latchkey runCheck work on a fresh VMAgents run the test suite the way CI will, on a fresh Linux VM, and get the real exit code back before they push.
latchkey watchPick up red CI automaticallyFailures self-heal recorded but did not fix go straight to your coding agent, with what self-heal recorded.
MCP serverWorks in any MCP clientClaude Code, Cursor and Codex connect with one key to list failed runs, pull failure bundles, tail run logs and run jobs.

Self-heal fixes transient failures mid-run, so agents only see the ones it did not fix.

More on pricing