Claude Code E2B Sandbox FastAPI Example

A FastAPI application that provides a REST API for running Claude Code in E2B sandboxes. This allows you to execute Claude Code commands remotely through HTTP requests, with support for session management and GitHub integration.

Features

Remote Claude Code Execution: Run Claude Code commands through HTTP API endpoints
Session Management: Resume conversations across multiple requests
GitHub Integration: Clone and push repositories to work with code in sandboxes
E2B Sandbox Environment: Isolated, secure execution environment
MCP (Model Context Protocol) Support: Enhanced capabilities through MCP servers

Prerequisites

Python 3.12 or higher
E2B account and API key
Anthropic API key (for Claude Code)
GitHub Personal Access Token (optional, for repository push)
Context7 API key (optional, for MCP Context7 server)

Installation

Clone the repository:

git clone https://github.com/e2b-dev/claude-code-fastapi
cd claude-code-fastapi

Install dependencies using uv (recommended):

uv sync

Or using pip:

pip install -e .

Environment Variables

Create a .env file in the project root with the following variables:

# Required
ANTHROPIC_API_KEY=your_anthropic_api_key_here
E2B_API_KEY=your_e2b_api_key_here

# Optional
E2B_SANDBOX_TEMPLATE=claude-code-dev  # or claude-code for production
GITHUB_PAT=your_github_personal_access_token_here
CONTEXT7_API_KEY=your_context7_api_key_here  # for MCP Context7 server

Environment Variable Propagation: The following environment variables are automatically passed to each E2B sandbox:

GITHUB_PAT - For GitHub repository access
CONTEXT7_API_KEY - For MCP Context7 server functionality
ANTHROPIC_API_KEY - For Claude Code execution

Building Templates

Development Template

Build the development template:

cd template
python build_dev.py

This creates a template with:

2 CPU core
4096 MB memory
Alias: claude-code-dev

Production Template

Build the production template:

cd template
python build.py

This creates a template with:

2 CPU core
4096 MB memory
Alias: claude-code

Running the Application

Start the FastAPI server:

uvicorn app.main:app --reload

The API will be available at http://localhost:8000

Access the interactive API documentation:

http://localhost:8000/docs

API Usage

POST /chat

Start a new conversation with Claude Code:

curl -X POST "http://localhost:8000/chat" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Can you add GPT 3.5-Turbo to the list of models and open a pull request on GitHub",
    "repo": "https://github.com/e2b-dev/fragments"
  }'

POST /chat/{session}

Resume an existing conversation using a session ID:

curl -X POST "http://localhost:8000/chat/038b769b-4717-47ca-be02-2a49bd7da978" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Now modify the script to accept a name parameter"
  }'

Request Body

{
  "prompt": "string", // Required: The prompt/command for Claude Code
  "repo": "string" // Optional: GitHub repository URL to clone (only for new sessions)
}

Note: When using /chat/{session}, the repo parameter is ignored as the repository is already cloned in the existing sandbox session.

Response

{
  "type": "result",
  "subtype": "success",
  "is_error": false,
  "duration_ms": 216401,
  "duration_api_ms": 216234,
  "num_turns": 81,
  "result": "Perfect! I have successfully completed all the tasks:\n\n✅ **All tasks completed!** The pull request has been created at: https://github.com/e2b-dev/fragments/pull/170\n\n## Summary\n\nI've successfully added GPT-3.5 Turbo to the model list and opened a pull request with the following changes:\n\n- **Added GPT-3.5 Turbo** to `/home/user/fragments/lib/models.json:108-114`\n- **Model configuration**:\n  - ID: `gpt-3.5-turbo`\n  - Provider: OpenAI\n  - Name: GPT-3.5 Turbo  \n  - Multimodal: false\n- **Validated** JSON syntax is correct\n- **Created branch** `add-gpt-3.5-turbo` \n- **Opened PR #170** with comprehensive description\n\nThe pull request is ready for review and includes all the necessary changes to integrate GPT-3.5 Turbo into the fragments application.",
  "session_id": "038b769b-4717-47ca-be02-2a49bd7da978",
  "total_cost_usd": 1.1459241,
  "usage": {
    "input_tokens": 300,
    "cache_creation_input_tokens": 77458,
    "cache_read_input_tokens": 2087724,
    "output_tokens": 13935,
    "server_tool_use": {
      "web_search_requests": 0
    },
    "service_tier": "standard"
  },
  "permission_denials": []
}

MCP Servers

You can add your own MCP servers to Claude Code by editing the template/.mcp.json file.

{
  "mcpServers": {
    "context7": {
      "type": "http",
      "url": "https://mcp.context7.com/mcp",
      "headers": {
        "Authorization": "Bearer ${CONTEXT7_API_KEY}"
      }
    }
  }
}