Claude Parallel Tasks MCP Server

How to use

You interact with the Claude parallel MCP server by configuring your MCP client to launch the local MCP endpoint and then sending requests to run several Claude prompts at once. Each task can include optional file contexts to enhance Claude’s understanding of the prompt, and the results can be redirected to separate files for straightforward review.

How to install

# Prerequisites
node -v
pnpm -v

# 1. Install Node.js and pnpm if not already installed
# (follow your platform-specific installation guides)

# 2. Clone the project and navigate into the directory
# (obtain this project from its source location)

# 3. Install dependencies
pnpm install

# 4. Set up environment variables
cp .env.example .env.local

# 5. Edit the local env file and add your Anthropic API key
# (open .env.local in your editor and set the key)
ANTHROPIC_API_KEY=your_anthropic_api_key_here

# 6. Build the project
pnpm run build

# 7. Run the server (stdio config shown below describes how to start locally)
pnpm start

To enable a local, runnable configuration, you can use a standard runtime command that launches the MCP server from its built distribution. The typical approach is to execute the server via Node, pointing to the built index file.

Configuration and startup notes

Configure your MCP client to connect to the local server, or run the server as a stdio endpoint using the following runtime command setup. The recommended local invocation shown here attains a running server that your MCP client can talk to.

{
  "mcpServers": {
    "claude_parallel": {
      "command": "node",
      "args": ["/path/to/claude-code-mcp/dist/index.js"],
      "env": {
        "ANTHROPIC_API_KEY": "your_anthropic_api_key_here"
      }
    }
  }
}

Security and environment

API keys are kept in environment variables rather than in source code, and the project’s local environment file is ignored by version control. Ensure you keep your Anthropic API key secure and do not share your local env files publicly.

Notes and usage patterns

- Use the run_parallel_claude_tasks tool to execute multiple prompts simultaneously with optional file contexts. Each query can specify contextFilePaths to supply additional materials for Claude to reference.

- Direct output for each task can be redirected to individual files to simplify post-processing and auditing.

run_parallel_claude_tasks

Executes multiple Claude prompts in parallel with optional file contexts. Each query can include a prompt and optional contextFilePaths to provide files as context.