buildkite-cli

name	buildkite-cli
description	Buildkite CLI for inspecting builds and parsing job logs. Use when the user shares a Buildkite build URL, wants to list jobs in a build, analyze CI test failures, asks about failing tests in a Buildkite job, or wants to check PR CI status from the current branch.

Buildkite CLI (bk)

The bk CLI inspects Buildkite builds and parses job logs. It can list all jobs in a build with pass/fail status, or download and parse individual job logs into structured output.

CLI availability

Run command -v bk before use. If it is unavailable, install the local source with cargo install --path ~/code/buildkite-cli.

Authentication

Set the Buildkite API token (requires read_builds, write_builds, and read_artifacts scopes):

export BUILDKITE_TOKEN="your-token"

Tokens can be created at: https://buildkite.com/user/api-access-tokens

Commands

bk pr checks [--branch <BRANCH>] [--compact/-c]

Shows Buildkite check status for a PR. Defaults to the current branch's PR; use --branch to check a different branch. Requires gh CLI to be authenticated.

When BUILDKITE_TOKEN is set, pending checks are enriched with actual Buildkite job states. This detects jobs that are failing behind a "pending" GitHub status (e.g. jobs in a retry loop that GitHub hasn't marked as failed yet).

Checks are sorted with failures first, then pending, then passed. Use --compact / -c for a denser format that puts URLs inline and drops redundant state labels.

bk pr checks
bk pr checks --compact
bk pr checks --branch darren/my-feature

Failed and pending checks show their Buildkite URL, which you can pass directly to bk jobs download-logs or bk builds list-jobs.

bk builds list-jobs <BUILD_URL> [--json]

Lists all jobs in a build with pass/fail status. Each job shows its full Buildkite URL (dimmed) on the next line for easy copy-paste into bk jobs download-logs.

bk builds list-jobs "https://buildkite.com/figma/ci/builds/287221"

Example output:

Build 287221 (ci)

  ✓ lint-check
    https://buildkite.com/figma/ci/builds/287221#aaa-bbb
  ✓ type-check
    https://buildkite.com/figma/ci/builds/287221#ccc-ddd
  ✗ multiplayer-rust-tests (failed)
    https://buildkite.com/figma/ci/builds/287221#eee-fff
  ✓ multiplayer-typescript-tests
    https://buildkite.com/figma/ci/builds/287221#ggg-hhh
  - deploy-staging (waiting)
    https://buildkite.com/figma/ci/builds/287221#iii-jjj

Use --json to output structured JSON with name, id, state, and url fields:

bk builds list-jobs "https://buildkite.com/figma/ci/builds/287221" --json

bk retry <JOB_URL>

Retries a specific failed job. The URL must include a #job-id fragment.

bk retry "https://buildkite.com/figma/figma/builds/5950766#019ca8a8-6e21-4548-9b5c-e8656a82feed"

bk jobs download-artifacts <JOB_URL> [--list] [--filter <GLOB>]

Downloads artifacts from a Buildkite job to the current directory. The URL must include a #job-id fragment. Useful when test output is stored as artifacts rather than inline in the log (e.g. Bazel test XML results).

# List artifacts without downloading
bk jobs download-artifacts --list "https://buildkite.com/figma/ci/builds/413652#019d6a74-..."

# Download all artifacts
bk jobs download-artifacts "https://buildkite.com/figma/ci/builds/413652#019d6a74-..."

# Download only XML test results
bk jobs download-artifacts --filter "**/*.xml" "https://buildkite.com/figma/ci/builds/413652#019d6a74-..."

bk jobs download-logs <JOB_URL>

Downloads and parses a job's log. The URL must include a #job-id fragment.

bk jobs download-logs "https://buildkite.com/figma/figma/builds/5950766#019ca8a8-6e21-4548-9b5c-e8656a82feed"

bk jobs download-logs --file <PATH>

Parses a local log file instead of fetching from the API.

bk jobs download-logs --file path/to/log.log --job-name multiplayer-rust-tests

Flags for jobs download-logs

Flag	Required	Description
<URL>	No*	Buildkite job URL with #job-id fragment
--file <path>	No*	Local log file to parse instead of fetching from API
--job-name	No	Job name hint when using --file (e.g. multiplayer-rust-tests)
--raw	No	Output cleaned log only, skip structured parsing

*Either a job URL or --file must be provided.

Output files

All files are written to --output-dir (default: current directory):

File	Description
{build}_{job}_raw.log	Original raw log
{build}_{job}_clean.log	Cleaned log (BK timestamps, ANSI, bazel noise removed)
{build}_{job}_results.json	Structured JSON with all test results
{build}_{job}_summary.txt	Human-readable summary with failures

Supported job types

multiplayer-rust-tests (nextest)

Parses nextest output across multiple bazel targets. For each nextest run, extracts:

• Per-test results (PASS, FAIL, TIMEOUT) with durations
• Failure details with captured stdout and stderr
• Run summaries (total, passed, failed, timed out, skipped)

multiplayer-typescript-tests (mocha)

Parses mocha test output. Extracts:

• Summary counts (passing, pending, failing)
• Each failure with suite name, test name, error message, diff, and stack trace
• Suppressed known flakes from the CI test-runner

agentplat-test and Go test jobs (gotest)

Parses Go test output run via Bazel. Matches job names containing agentplat or go-test. Extracts:

• Per-package results grouped by Bazel target
• Individual test pass/fail/skip with durations
• Failure output (error messages between === RUN and --- FAIL)
• Bazel summary (executed/total/passed/failed)
• Distinguishes executed vs cached targets (using Bazel progress line durations)
• Supports both Test output for (batch) and @@//target:binary | (streaming) Bazel formats
• Non-Go target failures (e.g. TS client itests): detects Bazel targets that fail without Go test output patterns. Captures raw error output and Bazel FAILED/TIMEOUT lines so these failures aren't silently hidden behind passing Go tests.

Generic script errors (fallback)

For any unrecognized job type, extracts error lines (ERROR:, TypeScript errors, Go compiler errors, ❌ markers, Bazel FAILED targets), the failed command, and exit code from the log. Go compiler errors (undefined:, too many arguments, etc.) include have/want type context lines.

When a specialized parser (gotest, golint, etc.) returns empty results (e.g. a compilation error prevents tests or linting from running), the CLI automatically falls back to this generic parser to surface the actual errors.

Lint jobs (golint)

Parses golangci-lint output. Matches job names containing lint. Extracts:

• Lint issues with file, line, column, message, and linter name

Examples

# Show Buildkite checks for the current branch's PR
bk pr checks

# Show Buildkite checks for a specific branch's PR
bk pr checks --branch darren/my-feature

# List all jobs in a build
bk builds list-jobs "https://buildkite.com/figma/ci/builds/287221"

# Download and parse a specific job's log
bk jobs download-logs "https://buildkite.com/figma/figma/builds/5950766#019ca8a8-6e21-4548-9b5c-e8656a82feed"

# Parse a previously downloaded log
bk jobs download-logs --file figma_build_5950766_multiplayer-rust-tests.log --job-name multiplayer-rust-tests

# Retry a specific failed job
bk retry "https://buildkite.com/figma/figma/builds/5950766#019ca8a8-6e21-4548-9b5c-e8656a82feed"

# List artifacts for a job
bk jobs download-artifacts --list "https://buildkite.com/figma/ci/builds/413652#019d6a74-..."

# Download test result artifacts
bk jobs download-artifacts --filter "**/*.xml" "https://buildkite.com/figma/ci/builds/413652#019d6a74-..."

# Just get the cleaned log for manual inspection
bk jobs download-logs --file build.log --raw | less

# Extract failing test names from JSON
cat *_results.json | jq '.runs[].failing_tests[].name'

Building

The CLI is a Rust project at /Users/dtsung/code/buildkite-cli:

cd /Users/dtsung/code/buildkite-cli
cargo build --release

The binary is at target/release/bk.