Log in
Check the deployed experienceBrowser QARun plain-language QA cases against a deployed preview in real Chrome, with per-case evidence, explicit uncertainty, and bounded AI spend.
ChangelogMore reviewers create more candidate findings, not necessarily more truth. Reliable multi-agent review needs stable identities, explicit dispositions, and evidence that survives rewording.
EngineeringHow Runnable reviews one exact pull request revision with bounded access, evidence-backed findings, explicit spend controls, and no repository code execution.
EngineeringTroubleshoot a GitHub Actions job stuck queued, understand what self-hosted runners can protect, and plan an independent CI fallback.
EngineeringFOLLOW THE BUILD
Release notes, control-plane internals, and the limits we would rather publish before you meet them in production.
FROM THE CONTROL PLANE
Run plain-language QA cases against a deployed preview in real Chrome, with per-case evidence, explicit uncertainty, and bounded AI spend.
ChangelogMore reviewers create more candidate findings, not necessarily more truth. Reliable multi-agent review needs stable identities, explicit dispositions, and evidence that survives rewording.
EngineeringHow Runnable reviews one exact pull request revision with bounded access, evidence-backed findings, explicit spend controls, and no repository code execution.
EngineeringTroubleshoot a GitHub Actions job stuck queued, understand what self-hosted runners can protect, and plan an independent CI fallback.
EngineeringA precise look at the workflow, action, runtime, and credential contracts behind Runnable compatibility reports.
EngineeringA second execution path for the workflows you already have, with explicit compatibility and one clean machine per job.
ChangelogWhat Runnable separates from GitHub Actions, what still depends on the source provider, and why that boundary matters.
EngineeringFrom envelope encryption to scoped release and log masking, the lifecycle of a secret inside a Runnable job.
EngineeringWhy untrusted workflow changes, customer secrets, write tokens, OIDC, and artifacts need a different policy on forked contributions.
EngineeringThe two Entire connection models, their event and identity boundaries, and how a job checks out the requested revision.
ChangelogHow Runnable's MCP tools expose bounded run state, masked logs, waiting, and controlled mutations for agent-driven operations.
EngineeringWhy Runnable reports unsupported workflow behavior before migration instead of discovering it during a release.
PrinciplesThe reproducibility, security, and lifecycle reasons every Runnable attempt gets a newly provisioned machine.
PrinciplesThe isolation model behind Runnable runners, job-scoped credentials, and logs that do not outlive their secrets.
EngineeringHow Runnable separates elapsed machine time, runner size, included usage, and spend controls.
Engineering