Chainguard Launches Secure Workspaces and Checks for AI Coding Agents

Supply chain security provider Chainguard launched two beta products, Chainguard Workspaces and Chainguard Checks, aimed at isolating untrusted code executions driven by developers and AI agents. Announced at the Assemble London event, the twin releases tackle security risks stemming from autonomous tools that require broad shell access, file editing capabilities, and package installation privileges. Rather than running generated code directly on developer laptops where corporate credentials reside, or routing small code iterations into congested build queues, the vendor built remote hardened execution environments, detailed in InfoQ's report.

This report covers the architecture, networking controls, developer integrations, and enterprise constraints for teams rolling out autonomous coding tools. Engineering leads and platform teams evaluating secure sandboxes can examine how microVM isolation pairs with local command synchronization, continuous verification, and signed build attestations. Alongside the two core betas, the rollout also spans the general availability of Chainguard Actions and early access availability for Chainguard Libraries across Go and .NET ecosystems.

Chainguard Secure Workspaces and Checks: the announcement

At its Assemble London event, Chainguard introduced Workspaces and Checks to secure the inner development loop from risks introduced by automated coding agents. Workspaces delivers dedicated remote microVMs designed to execute generated code without exposing local filesystems or corporate tokens. Checks shifts continuous integration into the active authoring cycle, using pre-warmed microVMs populated with curated packages to validate code edits before commits reach upstream repositories.

The announcement also formalized the general availability of Chainguard Actions, delivering a curated catalog of over 1,000 rebuilt and continuously maintained GitHub Actions. Simultaneously, Chainguard Libraries for Go and .NET entered early access, scanning incoming dependencies for malware or greyware while holding new packages behind a configurable availability delay. Chainguard shared that its beta scanner assessed over 5,000 packages across both ecosystems, discovering that over 10% carried risks with no matching records across existing public sources.

What actually changed with Chainguard Secure Workspaces and Checks

Under the new workflow, engineers invoke the chainctl develop command to synchronize a local working tree and uncommitted changes into a remote microVM. When a user detaches, the platform creates an environment snapshot that preserves file modifications, processes, and active shell contexts. Repositories can spin up multiple parallel Workspaces, allowing distinct agents to explore separate implementation paths without port collisions, state pollution, or shared caches.

Network security shifts away from embedding secrets inside guest instances. A host-side proxy injects short-lived credentials into approved outbound requests, while egress policy rules restrict and log connections attempting to reach unapproved destinations. When Checks runs alongside Workspaces, the workspace instance suspends during test execution and ingest structured logs if failures occur. Passing runs generate signed attestations linking the environment state, test outcome, and commit SHA directly to bypass redundant runs at the pull request merge gate.

Who should care about Chainguard Secure Workspaces and Checks

Platform security directors and DevOps leads managing developer environments benefit directly from keeping autonomous agents off local hardware. Because coding agents regularly install unvetted libraries and alter runtime settings, sandboxing them prevents prompt injection attacks or poisoned dependencies from accessing internal network segments or credentials. Moving execution to dedicated microVMs eliminates risks without requiring engineers to manually review every single intermediate command during generation cycles.

Teams wrestling with slow CI feedback loops also gain an immediate efficiency path. Instead of waiting for centralized CI runners to download dependencies repeatedly for minor patches, agents and developers can trigger Checks directly via a command-line interface, a pre-commit hook, an MCP server, an SDK, or a GitHub App. The architecture lets automated tooling test, fail, and fix syntax or test failures continuously within seconds, avoiding deep queue backlogs across shared infrastructure.

How to try Chainguard Secure Workspaces and Checks

Both Workspaces and Checks are currently accessible under 30-day beta trials. Teams can register to configure initial test environments and connect their workflows through the chainctl CLI or integrated GitHub App. Compute for Workspaces is billed based on active consumption, allowing organizations to pilot sandbox environments on representative projects before establishing full fleet allocations.

Current integrations in Workspaces focus on Anthropic's Claude agent, with support for additional agent frameworks planned for subsequent milestones. Integration hooks allow developers to trigger Checks from local terminal environments or configure automated workflows using the available MCP server. Platform teams can also explore the 1,000 rebuilt GitHub Actions catalog or enroll in the early access program for Go and .NET dependency scanning.

What to watch after Chainguard Secure Workspaces and Checks

Enterprise adoption will hinge on how effectively Chainguard proves its claims against seasoned sandbox and security incumbents. Sandboxing alternatives like E2B already run agent sessions inside Firecracker microVMs with egress controls, while StepSecurity Harden-Runner monitors runtime processes and network calls inside GitHub Actions runners. Meanwhile, dependency gates like Sonatype Repository Firewall and JFrog Curation already provide established quarantine and age-verification mechanisms across enterprise registries.

Prospective users must also weigh specific architectural constraints that Chainguard has yet to fully document. The vendor has not published comparative performance benchmarks showing execution speeds against traditional CI platforms, nor has it released verification guidelines detailing how customers independently audit its tamper-proof command logs. Furthermore, the efficiency of the merge gate relies completely on repositories trusting Chainguard's signing pipeline and guaranteeing that input parameters remain unmodified.