Skip to content
MCP server

ParallelSandbox

Give a coding agent a remote Linux box with Docker, a browser and human takeover

Pricing
Freemium
Platforms
API, Web, macOS
Runs
Locally or Remotely
Version
1.0.20
Listing updated
3 days ago

About ParallelSandbox

ParallelSandbox is a freemium MCP server for developer tools. It can run locally on your machine or as a hosted remote server.

ParallelSandbox treats a coding agent's finished work as something a person should open, not read. The agent starts a box with sandbox_start, puts code in with git clone or sandbox_sync for an uncommitted checkout, runs docker compose up and tests the result itself on a virtual display. When it has a version worth your time, sandbox_review pushes a card to the mobile or desktop app with a button that opens the running product on its own URL.

Each box is a full amd64 Linux machine with its own Docker daemon, 2 vCPU, 8 GB of memory and 40 GB of disk per size unit (1 to 4 units), restored from a snapshot in seconds. It has an Xvfb display with Chromium for headed tests, screenshots, mp4 recordings and a live view, and browser errors come back through logs_errors with stacks resolved through source maps. When a step needs a login, second factor or payment, the agent asks you to take over the screen.

The site says there is no AI inside the product. Usage is billed in credits: a free account gets 1,000 credits, with Pro at $10.99 and Max at $29.99 a month. The product is in beta.

Install

Run in a terminal

claude mcp add --transport http parallelsandbox https://mcp.parallelsandbox.com/mcp

Add to Settings → Connectors → Add custom connector

https://mcp.parallelsandbox.com/mcp
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "parallelsandbox": {
      "url": "https://mcp.parallelsandbox.com/mcp"
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "parallelsandbox": {
      "type": "http",
      "url": "https://mcp.parallelsandbox.com/mcp"
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "parallelsandbox": {
      "serverUrl": "https://mcp.parallelsandbox.com/mcp"
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "parallelsandbox": {
      "type": "http",
      "url": "https://mcp.parallelsandbox.com/mcp"
    }
  }
}

Configuration

  • PARALLELSANDBOX_API_KEYOptional existing psbx_ API key; without it the adapter opens OAuth sign-in
  • PSBX_TOOL_TIMEOUT_SECConfigured MCP host absolute tool timeout in seconds (default 60 gives a safe 25-second review wait; set both the host timeout and this value to 1920 for a 30-minute wait). Requires adapter restart.

Tools this server provides (6)

Taken from the project documentation; the running server may differ.

sandbox_start
Starts a box for a goal and the services your repo exposes.
sandbox_sync
Uploads an uncommitted working directory into a box.
sandbox_pull
Writes files from the box back to your machine.
sandbox_wire
Connects unchanged services living at an external base URL.
sandbox_review
Shows a version in the app and waits for feedback.
logs_errors
Returns browser errors with stacks resolved through source maps.

What it does

Full Linux machine
Real amd64 host with its own Docker daemon, not a container.
Visible browser
Xvfb display with Chromium for headed tests, screenshots and recordings.
Version handoff
Each version gets its own URL you open in your browser.
Human takeover
Take over the live screen for logins, second factors or payments.
Source-mapped logs
`logs_errors` returns browser errors with stacks resolved through source maps.
Local file sync
Upload uncommitted work or pull files back with `sandbox_sync` and `sandbox_pull`.

Good for

  • Let an agent test a docker compose stack end to end
  • Review a running version of a feature instead of a diff
  • Complete a login or payment step by taking over the box
  • Run only changed services in a box and wire them to your own dev environment

Pricing

Freemium, from $10.99 a month

Free 1,000 credits; Pro $10.99/mo (5,000 credits); Max $29.99/mo (20,000); credit packs

Prices change. Check parallelsandbox.com before you buy.

ParallelSandbox FAQ

What does a box contain?

An amd64 Linux machine with 2 vCPU, 8 GB memory and 40 GB disk per size unit, Docker, and Chromium on a virtual display.

How is it priced?

Credits: Free gives 1,000 once, Pro is $10.99 a month for 5,000 and Max is $29.99 a month for 20,000.

What happens when credits run out?

No card is charged. Boxes that can be frozen are frozen for 24 hours and stop billing.

How do I sign in?

The stdio adapter opens browser OAuth; an existing API key also works.

Is ParallelSandbox free?

ParallelSandbox has a free plan, with paid plans for more usage or features. Paid plans start at $10.99 a month.

How do I add ParallelSandbox to Claude Code?

Run this command in your terminal: claude mcp add --transport http parallelsandbox https://mcp.parallelsandbox.com/mcp

What are the best alternatives to ParallelSandbox?

The closest alternatives to ParallelSandbox are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

Alternatives to ParallelSandbox

  1. REA

    MCP server

    CLI and MCP server that lets coding agents reverse engineer apps and binaries

    Runs locally
    • Open source
    • 464K GitHub stars
    • Developer tools
  2. Codebase Memory

    MCP server

    Local MCP server that builds a code knowledge graph for AI coding agents

    Runs locally
    • Open source
    • 48.7K GitHub stars
    • Knowledge and memory
  3. GitHub

    MCP server

    Official MCP server connecting AI assistants to GitHub repos, issues and workflows

    Not checked yet
    • Remote
    • Open source
    • 34.5K GitHub stars
    • Developer tools
  4. MCP toolkit giving coding agents symbol-level code retrieval and editing

    Runs locally
    • Open source
    • 31.5K GitHub stars
    • Developer tools
  5. XERJ

    MCP server

    Local search engine for AI agents with cited answers over your code and docs

    Runs locally
    • Open source
    • 19.2K GitHub stars
    • Knowledge and memory
  6. Ouroboros

    MCP server

    Spec-first layer that pins requirements and verifies results of AI coding agent work

    Runs locally
    • Open source
    • 6.4K GitHub stars
    • Developer tools