Docker review

Docker is a container development platform, not one hosted API or a single server. Docker Desktop brings containers and the development loop together on a developer’s machine; Docker’s other products cover agent sandboxes, AI governance and hardened images. The platform sits between source code and the environments where applications run: teams build images, use registries and CI, then run containers locally or elsewhere.

For AI agents, Docker adds microVM-isolated sessions with network and filesystem controls. Its governance console applies sandbox and MCP policy and records identity-bound audit logs. Hardened Images are minimal, signed images and MCP servers that Docker says are continuously patched and built with SLSA Level 3. The homepage says teams can keep their existing images, registries and CI, and describes Docker as OCI-compliant and platform-independent. Agent work can start locally and move to Cloud Sandboxes. That makes Docker an extension to a container stack, rather than a replacement for the application code or CI system.

The named integration surface is Docker’s CLI and Docker Desktop. The homepage shows commands including docker agent run billing-bot and sbx run claude, and says developers can use existing images, registries and CI. Docker Desktop is offered for Mac with Apple silicon or Intel, Windows with AMD64 or ARM64, and Linux. That gives teams a local workflow and a stated path to Cloud Sandboxes for longer-running tasks.

The homepage does not establish a general-purpose API contract, SDK languages or a language-specific integration. If your application needs to create sandboxes or manage policy programmatically, confirm API and SDK support in Docker’s docs before choosing an integration pattern. The homepage links to further material on Sandboxes, governance and hardened images, but does not describe their API or SDK methods. Docker says its platform is OCI-compliant, which supports using standard container images; verify the behavior of your specific CI, registry and runtime integrations in the vendor’s docs.

Docker Personal costs $0 and includes one user, one Docker Scout-enabled repository, 100 Docker Hub pulls per hour and one private Docker Hub repository. DHI Community is also free per developer, with hardened minimal images, verifiable SBOMs, SLSA Build L3 provenance and an Apache 2.0 open-source image catalog.

Docker Pro is $11 per user monthly or $9 per user per month billed annually. Team is $16 monthly or $15 annually per user; Business is $24 under either billing option. Those plans increase Docker Scout repositories, private repositories, pull access or organization controls, and include 100, 500 or 1,500 Testcontainers Cloud runtime minutes per month respectively. Team supports up to 100 users; Business has no user cap.

Additional Testcontainers Cloud runtime minutes cost $3 per 100 minutes and expire at subscription end; on-demand minutes cost $4 per 100. DHI Select starts at $5,000 per repository. DHI Enterprise, Extended Lifecycle Support, Premium Support and TAM Services require custom pricing or have no published price.

Docker states that its platform is OCI-compliant and platform-independent, which gives teams a standards-based route for moving container images between compatible tools. The DHI Community catalog is open source under Apache 2.0. That licence applies to the catalog of open-source images, not automatically to Docker’s other products. The homepage does not establish a licence for Docker Desktop, AI Governance or Sandboxes.

Docker Desktop runs on a developer’s computer, and Docker says agents can run locally or in Cloud Sandboxes. That is not the same as a documented, self-hosted governance service. The vendor does not publish a data-export process on the homepage; ask how to export audit logs, policies and sandbox configuration before depending on them. OCI images can reduce runtime lock-in, but moving governance rules, logs or cloud sandbox workflows may require product-specific changes.

Docker’s production case for agents rests on microVM isolation, network and filesystem controls, signed components, policy enforcement and identity-bound audit logs. Its homepage also describes Hardened Images as continuously patched and SLSA Level 3. Those are relevant controls, but teams should still check the exact isolation boundary, policy behavior, image-update process and compliance evidence for the product and plan they intend to deploy.

The pricing page’s Testcontainers Cloud runtime-minute allowances are finite on Pro, Team and Business; additional and on-demand minutes have separate prices. Confirm expected consumption and the behavior at the limit before making runtime minutes part of a production pipeline. The homepage does not publish service reliability targets or general rate limits, so ask Docker about those and support coverage for your workload.

Developers on Reddit praise containers for reproducible environments and easier issue reproduction. Other commenters warn that containerizing every component can make small deployments clunky, and that unofficial images can carry security risk. Treat images and network access as production dependencies, not just setup details.