Overview
The Parallel Task MCP enables initiating deep research tasks or grouped tasks directly from an LLM client through the MCP interface. It serves as a convenient entry point to explore Parallel's APIs by running small experiments and composing production-ready workflows that depend on Parallel services. The MCP acts as a proxy to the central MCP endpoint at https://task-mcp.parallel.ai/mcp, facilitating access without building custom integrations from scratch. This repository provides a local testing workflow including wrangler dev, the Model Context Protocol inspector, and guidance to connect to the local proxy at http://localhost:8787/mcp. Configuration examples show how to register an MCP server named 'Parallel Task MCP' with its URL, illustrating how clients discover and route requests. The server is suited for developers who want to prototype, test, and illustrate MCP-powered task orchestration and API exploration within LLM-driven applications, while referring to the official MCP docs for more details.
Features
LLM-initiated task orchestration
Start and manage deep research tasks or task groups directly from an LLM client via MCP.
Batch task/group support
Coordinate and execute grouped tasks as a batch through the MCP server.
Proxy to hosted MCP endpoint
Provides a proxy to the MCP at task-mcp.parallel.ai/mcp for easy access.
Local development workflow
Supports local testing with wrangler dev and the Model Context Protocol Inspector; connect to http://localhost:8787/mcp.
Configurable MCP server entry
Demonstrates MCP server configuration via JSON (mcpServers) for easy setup.
Who Is This For?
- Developers:Use to trigger deep research and task groups from LLMs and explore Parallel APIs.
- LLM engineers:Integrate with LLM clients to orchestrate parallel tasks and experiments.
- Prototype engineers:Prototype production workflows using Parallel APIs and MCP in a test environment.
