Skip to main content

Claude Code Sandbox

Cosmonic Desktop is a local AI sandbox for Claude Code: the code Claude Code writes is compiled to WebAssembly and runs on your own machine with no access to your files, network, or keys until you grant it. Two settings wire it up, one for launching Claude Code from Builder with the sandbox playbook and one for installing Desktop's MCP server into Claude Code; use either or both.

Sandbox Claude Code with Cosmonic Desktop​

1. Settings → Agents: launch Claude Code from Builder​

Open Settings → Agents. The Detected agents card lists every coding agent Desktop found by its home directory (~/.claude) or on your login shell's PATH, and says whether each one is launchable from Builder. In the Skills card, turn on Claude Code to install the cosmonic-sandbox skill at ~/.claude/skills/cosmonic-sandbox/. This is the same install the one-time Set up your coding agents dialog offers on first run; nothing is installed silently.

The skill is the sandbox playbook: it teaches Claude Code to compile what it generates to a WebAssembly component and deploy it into the local sandbox instead of running it raw on your host. Once it is installed, Builder can launch Claude Code directly: open Builder, pick Claude Code as the agent, describe what you want, and its output lands in a sandboxed workload you can inspect before it runs.

For Claude Code the skill is the sandbox instructions on their own; the full Cosmonic plugin—slash commands, MCP wiring, and the sandbox gate—installs separately with claude plugin install cosmonic@cosmonic.

2. Settings → MCP Server: install Desktop's MCP server into Claude Code​

Open Settings → MCP Server. Under Install in your AI client, turn on Claude Code: Desktop runs claude mcp add for you. This registers cosmonicd mcp serve as an MCP server in Claude Code's own configuration, and the cosmonic_* tools (observe, deploy, build) are available the next time you start it. Expand Instructions on the same card to copy the manual snippet instead.

The manual equivalent (using the daemon binary's path on your platform—macOS shown):

claude mcp add --scope user cosmonic -- "/Applications/Cosmonic Desktop.app/Contents/Resources/cosmonicd" mcp serve

3. Verify the connection​

claude mcp get cosmonic   # shows the server and its connection state

Run claude mcp get cosmonic; it should list the server as connected while Desktop is running.

Skills over MCP​

Desktop's MCP server also publishes its skill family—cosmonic-sandbox, cosmonic-go, cosmonic-nats, cosmonic-nats-tuning, and cosmonic-kafka—through the MCP Skills extension (io.modelcontextprotocol/skills). A client that supports Skills over MCP gets the playbooks the moment it connects, with nothing written to its skills directory, so the MCP registration above is enough on its own for those clients. The on-disk skill from Settings → Agents covers clients that read a skills directory but do not yet speak the extension, and it is what Builder relies on to launch Claude Code.

What Claude Code can do once connected​

Its full tool surface (observe, deploy, build) and the guardrails are covered in Connect Coding Agents; the tools work only while Cosmonic Desktop is running.

Frequently asked questions​

How do I sandbox Claude Code?

Install Cosmonic Desktop, then under Settings → Agents turn on Claude Code to install the cosmonic-sandbox skill, and under Settings → MCP Server register Desktop’s MCP server in Claude Code. From then on the code Claude Code writes is compiled to WebAssembly and runs in a local sandbox with no access to your files, network, or keys until you grant it.

Does Claude Code itself run inside the sandbox?

No. Claude Code keeps running in your terminal or IDE exactly as before. What changes is where its output executes: the programs and MCP servers it builds run as sandboxed WebAssembly workloads on your machine instead of directly on your host.

Do I need both the skill and the MCP server?

They do different jobs. The MCP server gives Claude Code the tools to build, deploy, and observe sandboxed workloads; the skill gives it the know-how to use them by default. Desktop also publishes its skills over the MCP Skills extension, so a client that supports Skills over MCP gets the playbook from the connection alone.

What can the code Claude Code writes reach inside the sandbox?

Nothing by default. Each workload starts with no filesystem, network, or credential access; you grant exactly the capabilities it needs in its manifest, and everything else stays denied. The same manifest deploys unchanged to Cosmonic Control on Kubernetes.

Is this the same as Claude Code’s built-in sandbox?

No. Claude Code’s own sandboxing restricts the Bash tool inside the agent session. Cosmonic Desktop sandboxes the programs and MCP servers the agent builds: they run as WebAssembly components with no filesystem, network, or credential access until you grant it, and the same manifest deploys unchanged to Cosmonic Control on Kubernetes. The two are complementary.

Next steps​

Last updated on by Liam Randall