Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsCagent is the former name of Docker Agent, Docker’s open-source framework for building and running teams of AI agents from configuration files. You describe agents, their models, instructions, tools and delegation relationships in YAML or HCL; Docker Agent then orchestrates their work. Docker Desktop 4.63 and later uses the name Docker Agent, while versions 4.49 through 4.62 called the feature cagent. Docker’s current documentation is the place to check for current installation and usage details.
What is Docker Cagent?
Docker describes Docker Agent as “a framework for building and running custom agent teams.” It is a general-purpose agent runtime, not just an assistant for Docker commands. You define the agents and how they should work in a configuration file, then run the team from a terminal.
The “low-code” aspect is that configuration replaces much of the glue code normally used to define agent roles, connect tools and route work. It does not mean the system is entirely no-code: you still need to configure models and credentials, write useful instructions, and decide what tools and agents may do.
What happened to the name Cagent?
Docker’s documentation says the feature was called cagent in Docker Desktop versions 4.49 through 4.62. It is included as Docker Agent in Docker Desktop 4.63 and later. The name Cagent remains useful when looking for older references, but Docker Agent is the current name in the documentation.
#1 Best Overall
How does an agent team work?
A configuration describes a root agent and, optionally, specialist agents that it can call for particular tasks. Each agent can have its own instructions, model, parameters, context and tools. The root agent can delegate work rather than trying to perform every part of a task itself.
For example, a team might have a root agent that coordinates a request, a documentation agent that searches project files, and a shell-enabled agent that runs approved commands. The configuration specifies those roles and delegation links; the agents’ actual performance still depends on the selected models, instructions, available context and permitted tools.
What tools can agents use?
Docker Agent supports built-in capabilities such as task delegation, todo lists and memory, as well as filesystem and shell toolsets. It can also connect to external services through MCP servers. Tool access should be chosen deliberately: an agent with shell or filesystem access can take actions beyond simply generating text.
Can teams be shared?
Docker says agent configurations can be pushed to or pulled from Docker Hub or another OCI-compatible registry. This lets teams distribute configuration artifacts in a familiar registry workflow; it does not make the configuration itself a container image or guarantee that another user has the same credentials, models or connected services.
How do you set up and run Docker Agent?
Docker’s setup guide documents provider configuration, a setup wizard and a diagnostic command. A typical first run follows this order:
- Install Docker Agent. Docker Desktop 4.63 and later includes it. Docker also documents standalone routes for Docker Engine or custom installations, including Homebrew (
brew install docker-agent), Winget (winget install Docker.Agent), prebuilt binaries and source installation. If using it as a Docker CLI plugin, place the plugin in~/.docker/cli-pluginsand invoke it asdocker agent. Check the current Docker Agent page for version-specific installation instructions. - Choose and configure a model route. Use the
docker agent setupwizard or configure a supported provider, local model, custom endpoint or CLI harness as described in Docker’s setup guide. - Write an agent configuration. Define a root agent’s role, instructions and model in YAML or HCL. Add tools or specialist sub-agents only when the task needs them.
- Check readiness. Run
docker agent doctorto check provider credentials, local model availability and model auto-selection. Docker says the diagnostic does not print secret values. - Run the team. Start it with
docker agent run <agent-file>, replacing<agent-file>with the path to your configuration.
Docker packages and version support can change, so use its current documentation if a command or installation route does not match your environment.
Which model setup should you choose?
The choice affects cost, privacy, setup and hardware requirements. Docker’s setup documentation describes four broad routes:
| Route | Cost and prompt handling | What to consider |
|---|---|---|
| Hosted cloud provider | Generally billed per token; prompts are sent to the provider. | Configure provider credentials and review the provider’s data handling and pricing for your account. |
| Docker Model Runner (DMR) | Docker says local inference has no API key or per-token cost, and prompts stay on the user’s machine. | Download a model and ensure it fits available memory. Local hardware, storage, electricity and other services can still have costs. |
| Custom OpenAI-compatible endpoint | Depends on the endpoint and its operator. | Supply a base URL, API format and, where required, an environment variable for the key. Examples include vLLM, LiteLLM and corporate gateways. |
| Claude Code harness | Uses the separate Claude CLI and its subscription authentication, rather than a direct model-provider integration. | Docker’s documentation says the CLI bypasses permission prompts in non-interactive use; Docker advises using this route only in a trusted repository. |
Docker’s phrase “no per-token cost” for DMR concerns model inference after obtaining and running a local model; it does not mean local execution has no operating costs. Likewise, “prompts never leave your computer” describes the local-model route, not hosted providers, custom endpoints or every tool an agent may contact.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallAre Docker Agent and other Docker AI products the same?
No. Docker has several distinct AI-related products. Docker’s product documentation describes their roles as follows:
- Docker Agent configures and runs teams of agents.
- Gordon is Docker’s built-in assistant for Docker tasks such as debugging containers and writing Dockerfiles.
- Docker Model Runner runs models locally; Docker Agent can use it as a model route.
- MCP Catalog and Toolkit manage connections to external services through MCP.
- Docker Sandboxes provide an isolation layer for coding agents.
- Docker Agentic Platform is a separate experimental managed cloud service for running agents in Docker-managed cloud sandboxes. Docker describes its cloud compute as subscription-activated and pay-as-you-go.
These products can fit into related workflows, but Docker Agent and Docker Agentic Platform are not interchangeable: one is the agent framework, while the other is an experimental managed service.
What hardware does a local Docker agent need?
There is no single hardware minimum for Docker Agent because it can use hosted providers and different local models or configurations. One specific Docker Compose tutorial for an agentic AI sample asks for Docker Desktop 4.43 or later, Docker Model Runner enabled, at least 3.5 GB of VRAM and 2.31 GB of storage. Its example uses Gemma 3 4B with a context size of 10,000; the guide notes that a larger context configuration may use 7.6 GB of VRAM. Those figures apply to that tutorial’s sample stack, not to Docker Agent generally.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.When is Docker Agent a good fit?
Docker Agent is worth considering when you want a configurable team of agents, need to assign specialized roles or tools, or want to share agent configurations through an OCI-compatible registry. It offers flexibility across hosted models, local inference and compatible endpoints, but that flexibility means the setup requires decisions about credentials, privacy, model capability and tool permissions.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
If you only need help with routine Docker tasks, Gordon is the more directly focused assistant. If your priority is keeping prompts local, Docker Model Runner is the relevant model-execution option, but the model must fit your hardware. If you want Docker-managed cloud execution, Docker Agentic Platform is a separate experimental product rather than a setting inside Docker Agent.
What is established about Cagent’s history?
Docker’s current documentation establishes the product’s present name and the Docker Desktop version range that used the earlier cagent name. A Docker blog post by Stan Hamara, published November 13, 2025, described cagent’s Docker Desktop bundling and Agent Client Protocol integration, using Zed as an editor example. That post provides historical context; current setup instructions and naming are in Docker’s documentation.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
