Releasesby IbisKeep your focus on failures within group steps on the build canvas
UpdateyesterdayUnverifiedAdded Oct 14, 2026
Failed groups on the build canvas now put more focus on child steps that need your attention. When a group is failing or has finished with failures, expanding it shows only its failed steps. Click show all to reveal the hidden steps. Collapse and reopen the group to focus on failures again. State filters, such as Passed, still reveal matching steps.
More Buildkite releases
Every Buildkite releaseRetry jobs with narrower API token permissions
Retry automation no longer needs the broader write_builds REST API scope. The new retry_builds scope allows a token to retry failed jobs in existing builds or individual eligible jobs, without granting access to the create-build or rebuild endpoints. When creating or editing an API access token, select Build Retries ( retry_builds ).
Catch pipeline YAML errors before your first build with Buildkite MCP
Your AI agent can now check the pipeline YAML it writes before you commit or upload it. The Buildkite MCP Server includes validate_pipeline , a read-only tool that checks YAML syntax and structure against Buildkite's pipeline schema, without uploading the pipeline or starting a build.
Track your Pro plan spend and get alerted when it reaches a threshold
If your organization is on a monthly Pro plan, you can now see what your next invoice is shaping up to be before the billing period ends. Set a spend alert to get an email when the estimate reaches an amount you choose, so a busy month of builds doesn't become a surprise invoice.
Shipped this week
Buildkite shipped 40 releases. Three themes. Diagnosis. Agents put blocking failures first and retry infrastructure failures. Desktop alerts and log summaries are more reliable. Compatibility. GitHub Actions workflows get hashFiles() in composite actions, Windows checkout fixes, and retries for temporary token failures. Control.
Buildkite MCP Server updates
Your agent can now do more of the work between a failed build and a green one. Here are the latest tool updates for the Buildkite MCP Server . See what changed since the last green build When a build fails, one of the first questions is what's different from the last one that passed. compare_builds answers it in a single call.
Monitor billable active-user usage with the Buildkite API
You can now monitor your organization's billable active-user usage programmatically and compare it with your contracted allowance. Send the count to observability tools such as Datadog to track usage over time, or include it in internal license, billing, and compliance reporting.
Weekly wrap: September 27 – October 3, 2026
What shipped this week Between September 27 and October 3, 2026, we published 4 changelog entries and 35 releases across 13 public repos. Three things run through them: more control and predictability for platform teams, fewer gaps for teams moving off GitHub Actions, and more of the platform agents can drive directly. Tightened how secrets are handled.
Get notified about new macOS base image releases
You can now subscribe to RSS feeds for macOS base image releases on Buildkite hosted agents, making it easier to keep up with changes to your build environments. For updates to stable base images used in production, add the stable RSS feed to your preferred RSS reader.
Also shipped on Oct 13, 2026
Buildkite in October 2026What's included:
- Three training methods: Supervised fine-tuning (SFT) , direct preference optimization (DPO) , and reinforcement learning with group relative policy optimization (GRPO) . Each method trains a low-rank adapter (LoRA) or the full model weights. Custom loss functions : Train with any objective you compute in PyTorch from the model's log-probabilities.
The AI.AGG function is generally available (GA)
The AI.AGG function is generally available (GA). You can use this function to semantically aggregate unstructured input data based on natural language instructions.
BigQuery generative AI functions now support the gemini-4-argon model
BigQuery generative AI functions now support the gemini-4-argon model. This model is generally available .
Gemini 4 Argon in CodeMender (Preview)
CodeMender now supports Gemini 4 Argon. To use it, pass --model gemini-4-argon to cm find , cm verify , or cm fix . Gemini 3.8 Flash remains the default model. For more information, see Specifying the model .
Gemini 4 Argon and Gemini 4 Argon Cyber Permissive
Gemini 4 Argon ( gemini-4-argon ) is available in General Availability (GA) . Gemini 4 Argon is Google's frontier model that excels at deep work in complex, real-world tasks across domains including coding, finance, legal, cybersecurity, and science.
On-demand state repartitioning is now generally available
On-demand state repartitioning is now generally available. You can resize the number of shuffle partitions for a stateful Structured Streaming query or Lakeflow pipeline without losing checkpoint state. See On-demand state repartitioning for stateful streaming queries .