Open-source CLI

Deterministic CI failure classification.

Faultline turns failed CI logs into evidence-based failure reports. Same log in, same result out. No AI in the build path.

Try the CLI Join Teams early access

Open source View on GitHub See example reports

bash

$ faultline analyze failed.log

failure_id:    dependency.lockfile_drift
confidence:    0.91
evidence:
  - npm ci rejected the lockfile
  - package.json and package-lock.json are out of sync
suggested_fix:
  - regenerate the lockfile
  - commit package.json and package-lock.json together

Works with any CI runner that produces a build log

GitHub Actions GitLab CI CircleCI Jenkins Buildkite Drone CI TeamCity Semaphore

Why Faultline

Built for reproducibility, not exploration.

Faultline matches known failure classes against log evidence. Every result is traceable and stable across runs.

Deterministic

Same log produces the same failure classification every time. Results do not drift between runs or environments.

Explainable

Every classification links to specific evidence lines extracted from the log. You can audit exactly why a class was chosen.

Automation-safe

Structured JSON output with stable failure IDs. Safe to pipe into scripts, CI steps, Slack, or downstream tooling.

Output

Structured output, ready for automation.

Pass --json to get machine-readable output. Pipe it into your alerting, triage scripts, or the Teams sync.

Stable failure_id for deduplication across builds.
Confidence score and matched evidence lines included.
Suggested fix steps alongside every classification.
Works offline. No API calls during analysis.

faultline analyze failed.log --json

{
  "failure_id": "dependency.lockfile_drift",
  "confidence": 0.91,
  "evidence": [
    "npm ci rejected the lockfile",
    "package.json out of sync"
  ],
  "suggested_fix": [
    "regenerate the lockfile",
    "commit both files together"
  ]
}

Approach

Not another AI CI bot.

Faultline does not guess. It matches known failure classes, shows the evidence, and returns stable structured output. AI can be useful for exploration, but CI classification should be reproducible. The same log should produce the same result on Tuesday as it did on Friday.

Start classifying failures in minutes.

Install the CLI, point it at a failed build log, and get a structured failure report. No account required.

Try the CLI Teams early access