LABSCO SUMMARY

A hosted endpoint you add in one line.

Because it is a URL rather than a package, trying it takes a config entry and a Parallel sign-in — which is what makes it a reasonable way to find out what the Task API can do before you write code against it. Self-hosting gets you the same endpoint behind your own worker.

What it is

A hosted MCP endpoint for Parallel's Task API: start deep research runs or task groups from the client you already use, without writing against the API first.

What you get

Deep research started from a prompt in your own client
Task groups initiated the same way
A way to learn what Parallel's APIs do by using them, and to run small experiments while building production systems on top of them

Requirements

No install, but it needs a Parallel account: add https://task-mcp.parallel.ai/mcp to your client, then sign in with OAuth or send an API key from platform.parallel.ai as a Bearer token. The repository is a proxy to that hosted endpoint; running it locally is wrangler dev, which serves at http://localhost:8787/mcp.

Setup effort

Paste a URL, then authorize — add the endpoint to your client, then approve the OAuth consent screen

See installation options

USE CASES

01

Running a research task without building an integration

Add the URL to your client and start the run from the conversation you were already in.

02

Working out whether the API suits your problem

Try the shape of a task by asking for one, before committing engineering time to it.

03

Testing a change while you build on the API

Fire a one-off run from the client instead of writing a script for every experiment.

04

Keeping the endpoint under your own control

Run the proxy yourself with wrangler and point clients at your local address.

BEFORE YOU INSTALL

What to know first

Requires

A client that can reach a remote MCP server over HTTP

INSTALLATION

Connect your client

Add the endpoint to your client, then approve the OAuth consent screen.

Server setup

ENDPOINT URL

https://task-mcp.parallel.ai/mcp

MCP Clients

Claude Code

CLAUDE CODE · TERMINAL

claude mcp add --transport http "Parallel-Task-MCP" https://task-mcp.parallel.ai/mcp

Then run /mcp inside Claude Code and finish the sign-in in your browser.

Claude.ai

Go to Settings → Connectors → Add Custom Connector.

Name: Parallel Task MCP

URL:

CLAUDE.AI · URL

https://task-mcp.parallel.ai/mcp

Sign in to Parallel when the connector asks. In an organization you may need your administrator to add custom connectors.

Codex

CODEX · TERMINAL

codex mcp add parallel-task --url https://task-mcp.parallel.ai/mcp \
  --bearer-token-env-var PARALLEL_API_KEY

Set PARALLEL_API_KEY in your environment first, with a key from platform.parallel.ai. Leave off --bearer-token-env-var to sign in with OAuth instead. Restart Codex afterwards.

Cline

CLINE · JSON

{
  "mcpServers": {
    "Parallel Task MCP": {
      "type": "streamableHttp",
      "url": "https://task-mcp.parallel.ai/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_PARALLEL_API_KEY>"
      }
    }
  }
}

<YOUR_PARALLEL_API_KEY> is an API key from platform.parallel.ai. This entry signs in with the key, so there is no OAuth consent screen.

Parallel Task MCP