LoginSign Up
← Back to all sparks
B

Buildkite

INFRA · APIS
Velocity6.3

Buildkite widens its API surface for agent-driven CI debugging and observability

ci-cdrest-apiobservabilityhosted-agentsmcpsecurity
◆Current state
Buildkite's recent releases cluster around one theme: exposing more of the CI runtime through APIs. Richer REST job and agent objects, per-job performance metrics, and MCP server tooling all aim at automated and agent-driven consumers, alongside a security fix and an infrastructure notice.
◆Where it's heading
The platform is being reshaped for programmatic and agentic use — surfacing signal_reason and runner context so automation can tell infrastructure failures from code failures, adding job-level CPU/memory/disk metrics, and splitting jobs from builds for large-matrix querying. The MCP investment (elsewhere in the feed) is the same bet from another angle.
◆Prediction
Expect the REST and GraphQL surfaces to keep expanding toward machine consumers, with the MCP server becoming the primary interface for automated build triage.

◆Recent moves

  1. 2d ago

    Hosted Agent Outbound IP Address Ranges Updating August 2, 2026

    An operational heads-up: Hosted Agent outbound IP ranges expand on August 2, 2026. No product capability changes — only teams maintaining allowlists need to act.

  2. 10d ago

    Diagnose infrastructure-related job failures through the REST API

    Exposing signal and signal_reason plus runner lifecycle fields on REST objects gives automated debugging loops a way to distinguish infrastructure failures from code failures — a concrete step in the agent-facing API arc.

  3. 17d ago

    Read-only API tokens no longer expose pipeline webhook URLs

    Closing the read-only-token webhook_url leak is a security correctness fix: tokens now apply the more restrictive of scope and creator permissions. It's a breaking change for anyone reading webhook URLs with read-only tokens.

  4. 22d ago

    Improvements to log groups for running builds

    Live durations, sticky headers, and auto-expanding log groups make running builds easier to follow — a focused UX improvement on the new build page rather than a directional move.

  5. 22d ago

    Track the performance of your hosted agents

    A zero-config Metrics tab showing CPU, memory, and disk per job adds native observability to hosted agents, helping teams right-size instances and catch OOM failures without wiring up dashboards.

  6. 24d ago

    REST API job endpoints for large builds

    Fetching jobs separately from builds via exclude_jobs and dedicated job endpoints makes high-parallelism and large-matrix builds practical to query — another increment in the API-scalability push.