MCP server for E2B code sandbox - Execute code and commands in secure E2B sandboxes
Wellknown found it in public sources; nobody has proven control of it yet. Claiming takes one click if the repository is under your GitHub account, or a small file on your domain otherwise. Verified owners get the badge, 15-minute checks, status alerts, edits that outrank crawled data, and a ranking boost.
Agents can do it too: POST https://wellknown.network/api/v1/claims with {"agent":"code-sandbox-mcp","method":"well_known_file"} — machine-readable steps at claim.json, guide at /docs/claim.
Everything here was measured by our prober or read from a registry. Nothing is self-reported.
Attributed to the source that supplied each field. Treated as claims, not facts.
# Code Sandbox MCP MCP (Model Context Protocol) server for E2B code sandbox. This server provides tools to create, manage, and execute code in secure E2B sandboxes. ## Features - **Create Sandbox**: Create new E2B sandboxes with customizable timeout and template - **Kill Sandbox**: Terminate existing sandboxes by ID - **Run Code**: Execute Python code in sandboxes using Jupyter Notebook syntax - **Run Command**: Execute shell commands in sandboxes ## Installation ```bash pip install -e . ``` Or using uv: ```bash uv pip install -e . ``` ## Configuration Create a `.env` file in the project root with your E2B credentials: ```env E2B_API_KEY=your_api_key_here E2B_DOMAIN=your_domain_here ``` ## Usage ### Running the Server Using MCP CLI with Inspector (for testing): ```bash uv run mcp dev code_sandbox_mcp/server.py ``` Or using MCP CLI: ```bash uv run mcp run code_sandbox_mcp/server.py ``` Or directly with Python: ```bash python -m code_sandbox_mcp.server ``` Or using the installed command: ```bash code-sandbox-mcp ``` ### Available Tools 1. **create_sandbox(timeout: int = 300, template: str = None)** - Creates a new E2B sandbox - Returns the sandbox ID - Parameters: - `timeout`: Sandbox timeout in seconds (default: 300) - `template`: Optional template name or ID 2. **kill_sandbox(sandbox_id: str)** - Kills an existing sandbox - Returns empty string on success, error message on failure 3. **run_code(code: str, sandbox_id: str = None)** - Executes Python code in a sandbox - Uses Jupyter Notebook syntax - If `sandbox_id` is provided, connects to existing sandbox - Otherwise creates a new sandbox - Returns stdout and stderr 4. **run_command(command: str, sandbox_id: str = None)** - Executes a shell command in a sandbox - If `sandbox_id` is provided, connects to existing sandbox - Otherwise creates a new sandbox - Returns stdout, stderr, and exit code ## Development This project uses: - Python 3.…
Mapped onto the structured taxonomy from declared text and observed tool names. Confidence shown for derived entries.
Every source is kept verbatim. Field changes are logged as events.