LoginSign Up
← Back to all sparks
B

Buildkite

INFRA · APIS
Velocity8.8

Buildkite is rebuilding CI as something agents drive, not just humans.

ci-cdmcpagentsobservabilityintegrationstest-analytics
◆Current state
Buildkite's recent releases split in two: MCP tooling for agents - list_tests, wait_for_build, compliance with the stateless MCP spec - and broader REST/GraphQL coverage so setup, governance, and troubleshooting can run headless. Conventional CI work continues underneath, with checkout configuration in pipeline YAML and OpenTelemetry job log export. The framing on nearly every note is what an agent can now do without a human in the loop.
◆Where it's heading
Buildkite is arguing that CI should be forge-independent, and it is now backing that with coverage: GitHub, GitLab, Bitbucket, and as of this week Cursor's Origin, where it shipped as a launch partner on day one. On the agent side the pattern is removing the workarounds automation used to need - a wait tool instead of a polling loop, aggregated test metrics instead of assembling individual runs, an early failure signal instead of waiting for a job to exit. Each release converts a hand-rolled agent behavior into a first-class primitive.
◆Prediction
The read side of the MCP server is now largely covered, so expect write-side tools next - retrying jobs, unblocking builds, creating or editing pipelines from an agent - following the same pattern the REST expansion just established.

◆Recent moves

  1. 2d ago

    Buildkite is a Cursor Origin launch partner

    Buildkite ships in the Origin Marketplace the day Cursor's git forge launches, alongside its existing GitHub, GitLab, and Bitbucket support, with an option to keep GitHub as the source of truth and use an Origin mirror to cut cloning overhead across builds. It is the same forge-independent position Buildkite has held for years, applied to a forge that did not exist last week.

  2. 7d ago

    Buildkite MCP Server can now find your slowest and flakiest tests

    A read-only list_tests tool gives agents a suite-wide view of Test Engine tests with reliability and duration metrics, sortable and filterable by branch, owner, label, or time window. It is the MCP-side twin of the Test Engine API change shipped days earlier - the same data, exposed where an agent already is, so flaky-test triage does not start with assembling individual runs.

  3. 13d ago

    Analyze test reliability and performance with the Test Engine API

    The Test Engine list-tests endpoint now returns aggregated reliability, duration, and execution metrics per test, with filtering, relative or explicit time ranges, and sorting by reliability or duration. The change is gated behind a version header so existing integrations are untouched, which is how Buildkite has been shipping API changes lately.

  4. 13d ago

    Customize Git checkout behavior directly in pipeline YAML

    A checkout block in pipeline YAML now controls shallow clones, sparse paths, submodules, custom Git flags, or skipping checkout entirely, at pipeline or step level. It folds three separate plugins and a set of environment variables into the pipeline definition - housekeeping, but the kind that removes a class of configuration drift.

  5. 14d ago

    More Buildkite workflows are available through APIs

    Pipeline creation, organization administration, audit events, notification services, hosted agent resources, and artifacts all gain REST and GraphQL coverage. A broad sweep rather than a single feature, and the clearest statement of the headless-CI goal: anything an administrator does should be scriptable.

  6. 14d ago

    Send job logs to your OpenTelemetry collector

    Agent v3.135.0 can export job logs to an OpenTelemetry collector, correlated with job, phase, and hook spans when tracing is on, so CI output lands in the same backend as application telemetry. It fits the wider push to make build state readable by tools rather than only by people scrolling a log pane.