Exec Sandbox MCP server
Runs caller-supplied Python code in a network-isolated, locked-down sandbox and returns execution results.
Also known as inhuman/mcp-exec, mcp-exec
- Stars
- 1
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- 0.5.4
- Last commit
- Jul 20, 2026
- License
- MIT
What is Exec Sandbox MCP?
Exec Sandbox runs caller-supplied Python code in a network-isolated, locked-down sandbox and returns execution results.
Summary generated from the project's README
About Exec Sandbox MCP server
Exec Sandbox provides a secure environment for running Python code through a single tool called exec. It is designed for developers building AI agents that need to perform complex tasks by writing code rather than relying on hundreds of individual tool schemas. The server supports three different transport methods, including stdio, HTTP, and SSE, making it adaptable to various client setups.
Users can run code in a highly restricted environment that features a read-only rootfs, non-root execution, and strict limits on memory, CPU, and PIDs. The sandbox includes Python 3 along with common libraries like numpy, pandas, matplotlib, Pillow, Jinja2, and PyYAML. When using the stdio transport, the network is completely disabled, while HTTP and SSE users can implement additional security via an auth token or Kubernetes NetworkPolicies.
Execution results include stdout, stderr, exit codes, and duration. The system handles timeouts and caps output at 1 MiB to prevent resource exhaustion. All runs are serialized within a single instance, and ephemeral temporary directories are cleaned up after every task.
What you can do with it
- 1Perform secure code orchestrationInstead of using many small tools, an agent can write a single script to manage a multi-step workflow.
- 2Run isolated computationsDevelopers can execute untrusted code in a locked-down environment with no network access and limited resources.
Written from the project's README. Every use case is checked against a quote from it.
How to install Exec Sandbox MCP
Formats verified 10 Nov 2026Add to CursorAdd to GooseInstall Exec Sandbox MCP in Claude Code
$ claude mcp add --env 'MCP_EXEC_AUTH_TOKEN=<YOUR_MCP_EXEC_AUTH_TOKEN>' mcp-exec -- docker run -i --rm docker.io/idconstruct/mcp-exec:v0.5.4Install Exec Sandbox MCP in Cursor
~/.cursor/mcp.json
{
"mcpServers": {
"mcp-exec": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"docker.io/idconstruct/mcp-exec:v0.5.4"
],
"env": {
"MCP_EXEC_AUTH_TOKEN": "<YOUR_MCP_EXEC_AUTH_TOKEN>"
}
}
}
}Install Exec Sandbox MCP in VS Code
.vscode/mcp.json
{
"servers": {
"mcp-exec": {
"type": "stdio",
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"docker.io/idconstruct/mcp-exec:v0.5.4"
],
"env": {
"MCP_EXEC_AUTH_TOKEN": "<YOUR_MCP_EXEC_AUTH_TOKEN>"
}
}
}
}Install Exec Sandbox MCP in VS Code CLI
$ code --add-mcp '{"name":"mcp-exec","type":"stdio","command":"docker","args":["run","-i","--rm","docker.io/idconstruct/mcp-exec:v0.5.4"],"env":{"MCP_EXEC_AUTH_TOKEN":"<YOUR_MCP_EXEC_AUTH_TOKEN>"}}'Install Exec Sandbox MCP in Claude Desktop
claude_desktop_config.json
{
"mcpServers": {
"mcp-exec": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"docker.io/idconstruct/mcp-exec:v0.5.4"
],
"env": {
"MCP_EXEC_AUTH_TOKEN": "<YOUR_MCP_EXEC_AUTH_TOKEN>"
}
}
}
}Install Exec Sandbox MCP in Codex
$ codex mcp add mcp-exec --env 'MCP_EXEC_AUTH_TOKEN=<YOUR_MCP_EXEC_AUTH_TOKEN>' -- docker run -i --rm docker.io/idconstruct/mcp-exec:v0.5.4Install Exec Sandbox MCP in Codex config.toml
~/.codex/config.toml
[mcp_servers.mcp-exec] command = "docker" args = ["run", "-i", "--rm", "docker.io/idconstruct/mcp-exec:v0.5.4"] [mcp_servers.mcp-exec.env] MCP_EXEC_AUTH_TOKEN = "<YOUR_MCP_EXEC_AUTH_TOKEN>"
Install Exec Sandbox MCP in Gemini CLI
$ gemini mcp add -s user -e 'MCP_EXEC_AUTH_TOKEN=<YOUR_MCP_EXEC_AUTH_TOKEN>' mcp-exec docker -- run -i --rm docker.io/idconstruct/mcp-exec:v0.5.4Install Exec Sandbox MCP in Gemini settings.json
~/.gemini/settings.json
{
"mcpServers": {
"mcp-exec": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"docker.io/idconstruct/mcp-exec:v0.5.4"
],
"env": {
"MCP_EXEC_AUTH_TOKEN": "<YOUR_MCP_EXEC_AUTH_TOKEN>"
}
}
}
}Configuration
MCP_EXEC_AUTH_TOKENSecretOptional X-MCP-AUTH token (http/sse transports only).
MCP_EXEC_DEFAULT_TIMEOUT_SDefault wall-clock timeout per exec, seconds.
MCP_EXEC_MAX_OUTPUT_BYTESCap on combined stdout/stderr before truncation.
MCP_EXEC_MAX_STDIN_BYTESCap on stdin passed to the sandboxed code.
MCP_EXEC_MAX_TIMEOUT_SMaximum wall-clock timeout a caller may request, seconds.
MCP_EXEC_TRANSPORTMCP transport. Must be 'stdio' for direct docker/stdio use.
| Name | Description | |
|---|---|---|
| MCP_EXEC_AUTH_TOKEN | Optional X-MCP-AUTH token (http/sse transports only). | Secret |
| MCP_EXEC_DEFAULT_TIMEOUT_S | Default wall-clock timeout per exec, seconds. | |
| MCP_EXEC_MAX_OUTPUT_BYTES | Cap on combined stdout/stderr before truncation. | |
| MCP_EXEC_MAX_STDIN_BYTES | Cap on stdin passed to the sandboxed code. | |
| MCP_EXEC_MAX_TIMEOUT_S | Maximum wall-clock timeout a caller may request, seconds. | |
| MCP_EXEC_TRANSPORT | MCP transport. Must be 'stdio' for direct docker/stdio use. |
Automated checks
Ran Nov 15, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachableinhuman/mcp-exec is public and reachablegithub.com/inhuman/mcp-exec ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-07-20github.com/inhuman/mcp-exec ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubmcp-exec
English | Русский
Public OSS MCP server (Go, MIT) exposing a single powerful tool — exec — that runs
caller-supplied Python code in a network-isolated, locked-down sandbox and returns
stdout / stderr / exit_code. It is the "code-execution mode" building block: instead of
flooding an agent's context with hundreds of tool schemas, the agent writes code that orchestrates
the work.
Works over three transports — stdio / HTTP / SSE — with an identical tool set everywhere
(official modelcontextprotocol/go-sdk).
The exec tool
Input: { code: string (required), timeout_s?: int, stdin?: string }
Output: { stdout, stderr, exit_code, duration_ms, truncated, timed_out }
- A non-zero
exit_codeortimed_out=trueis a normal result, not a tool error. Only invalid input (emptycode, oversizedstdin) is a tool-call error. - Sandbox v1: Python 3 with stdlib + PyYAML, Jinja2, Pillow, numpy, pandas, matplotlib.
Security model (invariants)
Per execution: no network, non-root, cap-drop=ALL, no-new-privileges, read-only rootfs,
no CAP_SYS_ADMIN, ephemeral tmpdir (cleaned up), wall-clock timeout (kills the whole
process-group), memory/PID/CPU limits, capped output (1 MiB → truncated). Runs are serialized
within an instance; scale out with replicas. Caller data (code/stdin/output) is never persisted
and never logged in full — only metadata.
execis the most powerful surface there is. When embedding it in an agent, gate it behind that agent's tool-policy (trusted roles only).
Network note: --network none only applies to stdio. In HTTP/SSE the container needs
networking to serve its port, so sandboxed code would inherit egress — deny it at the orchestrator.
In k8s, a NetworkPolicy that allows ingress to the port and denies all egress:
apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata: { name: mcp-exec-lockdown }
spec:
podSelector: { matchLabels: { app: mcp-exec } }
policyTypes: [Ingress, Egress]
ingress: [{ ports: [{ port: 8080 }] }]
egress: [] # deny all egress → sandbox has no network
Run
go build -o mcp-exec ./cmd/mcp-exec
MCP_EXEC_TRANSPORT=stdio ./mcp-exec
Docker (recommended production posture):
docker run --rm -i --network none --read-only --cap-drop ALL \
--security-opt no-new-privileges --user 65532:65532 \
--tmpfs /tmp:rw,noexec,nosuid,size=64m \
--memory 256m --pids-limit 128 --cpus 1 \
idconstruct/mcp-exec
--tmpfs /tmp is required: the rootfs is read-only, and each run needs a writable
ephemeral workspace (cleaned up after).
Optional auth (HTTP/SSE)
Set MCP_EXEC_AUTH_TOKEN to require every HTTP/SSE request to carry a matching X-MCP-AUTH header
(constant-time compare; 401 otherwise). Empty token disables it. Not applicable to stdio.
Configuration
| Env var | Purpose | Default |
|---|---|---|
MCP_EXEC_TRANSPORT |
stdio | http | sse |
stdio |
MCP_EXEC_ADDR |
listen address for http/sse | :8080 |
MCP_EXEC_DEFAULT_TIMEOUT_S |
default wall-clock timeout | 30 |
MCP_EXEC_MAX_TIMEOUT_S |
timeout ceiling | 300 |
MCP_EXEC_MAX_OUTPUT_BYTES |
combined stdout+stderr cap | 1048576 |
MCP_EXEC_MAX_STDIN_BYTES |
stdin size cap | 1048576 |
MCP_EXEC_PYTHON |
interpreter path | python3 |
MCP_EXEC_AUTH_TOKEN |
if set, http/sse require X-MCP-AUTH header (constant-time); empty = off |
`` |
Not in v1
bash / multi-language, network from the sandbox, proxying other MCP servers into the sandbox.
License
MIT.
Exec Sandbox MCP questions
How do I add Exec Sandbox MCP to Claude Code?
Run claude mcp add --env 'MCP_EXEC_AUTH_TOKEN=<YOUR_MCP_EXEC_AUTH_TOKEN>' mcp-exec -- docker run -i --rm docker.io/idconstruct/mcp-exec:v0.5.4 in your terminal. Add --scope user to make it available in every project.
How do I add Exec Sandbox MCP to Cursor?
Open ~/.cursor/mcp.json and add the server under mcpServers, using the Cursor config shown in the install section above.
How can I secure HTTP or SSE connections?
You can set the MCP_EXEC_AUTH_TOKEN environment variable to require an X-MCP-AUTH header for all requests.
What are the supported transport methods?
The server works over stdio, HTTP, and SSE transports.