CI/CD pipelines compromise is hard to detect and supply chain attacks continue to rise. We’re investigating monitoring runtime behavioral signals to catch supply chain attacks proactively and ensure CI/CD tampering is immediately identified and managed.

Teams can't reliably answer what dependencies shipped in an artifact or prove the artifact came from their pipeline unmodified, introducing risk and audit failure. Autom generate and cryptographically attests SBOMs and provenance with zero configuration, leveraging mature tools (Syft, cosign, Sigstore).

When a single developer commits 30 times a day, it's not enough to know if the tests pass. We're investigating how to systematically validate change from a quality and security perspective beyond tests, without requiring intensive human intervention, so you can be confident in what you release in the ways that matter to your team and your org.

As we identify areas of struggle, where CI maintenance becomes a requirement for dependable CI greatness, we'll release new tasks. We move quickly, so check back often to see what Chunk can do next.

A series of improvements to hosted resource classes that will allow customers to benefit from increasingly powerful executors, ensuring the fastest feedback loops and maximizing productivity.

We’re investigating how to ensure the agent has just what it needs to do the most effective work fixing code with the greatest success rate and token efficiency. We plan to share the results of our experiments, so keep an eye out for more. Condensed step output gives agents a condensed summary of every failed step. You can also copy a ready-made diagnostic prompt straight from a failed workflow in the UI and paste it directly to your agent o

Enforceable, platform-level egress controls as a native feature.

A body of engineering work to make caches more performant and reliable, reducing time spent running build pipelines

Work is underway to evolve the familiar power and convenience of Orbs to make it ultra easy to use orbs, write your own orbs and debug orbs, for both humans and agents.

Chunk loves toil, because you have better things to do. Try our 5 initial tasks today and rest assured they're getting better with every use. We're making it possible to set it and forget and watch your CI improve.

So much testing insight is lost in the blob. We're bringing that insight to you in a thoughtful, actionable way. Don't just run tests, (especially when agents write them) understand them and their impact.

Chunk CLI is a command-line tool for running coding tasks on CircleCI’s infrastructure. Describe the task from your terminal, and Chunk sidecars carry it out where your pipelines already run, with your project’s stack and credentials in place. You keep working locally while the task runs, and the result comes back as a pull request you review like any other.

We've made self-hosted runner pools easier to operate: a single deployment can now scale multiple resource classes at once, complete with a ready-to-use pool VM image, flexible time-based scaling, and hot-reload of config and secrets. The latest release adds SSH-based debugging into KubeVirt VM runners, proxy support for restricted network environments, and simplified token management using per-resource-class credentials.

Thanks to our robust architecture and deep understanding, we're proud to offer unlimited job concurrency on all Linux resources to our scale customers at no additional cost. Never be blocked by your own productivity. Scale plans aren't for everyone, to understand if they're for you, contact us.

IP ranges are now supported for x86 Docker Gen2 resources classes. Pricing for IP ranges is unchanged and is the same across all executor generations.

We've gotten lots of feedback about the Slack orb. Based on that feedback we're proud to launch CircleCI’s native Slack integration - shared project notifications for the team, and personal DM notifications for you, no orbs required!

Optimized for agents like Claude Code and Codex, the new CLI includes a built-in MCP server that wraps the entire CLI. and gives you txhe ability to trigger and monitor pipeline runs, and access to build outputs. That means an agent can start a run, watch it, and read the failure output without polling leaving your terminal.

The hosted MCP server exposes a curated set of tools for observing and acting on CI runs: inspecting runs, reading logs, and rerunning or canceling workflows. CircleCI hosts it, so there's nothing to install locally. Your AI assistant connects over HTTPS and authenticates with OAuth 2.0 or a personal API token.

Introducing Smart Deployment. Validate every release against your monitoring signals and rolls it back automatically when one fails. Set it up in the UI, point your monitoring tool at CircleCI, and define the pass/fail policy that triggers your rollback pipeline.

See where your credits go on a per-project basis with our new Project Usage Dashboard, now available in Preview to all CircleCI Cloud users. Summary metrics like Credits per Pipeline, credits spend per user, usage trend and more, all at your fingertips.

Certificate manager, auto-renewal, and team sync — end certificate chaos permanently and get to your end user faster.

Chunk sidecars are microVMs hosted on CircleCI for coding agents to validate changes as they loop. Run microbuilds (pre-configured lightweight CI checks) in a sidecar while your agent is still working. Catch failures before the push. Learn more at chunk.ai

Some teams have a strong sense of how to get the most out of CI and others don't. Budget controls and alerting help you understand when teams' or projects' credit consumption suddenly increases, so you can ensure they are making the most of their validation system. Not available on all plan types.

Reduce test cycle times by up to 75% by intelligently running only the tests that matter for your specific code changes. Now in Beta.