Skip to content

Summary

A sandbox for third-party API integrations: agents replay Stripe, GitHub, Twilio and 50+ other workflows — webhooks, retries and failure cases included — without touching the real API.

Features

  • Replays full integration workflows: request, webhook, retry, state check
  • Chaos cases — duplicate, out-of-order and stale webhook deliveries
  • 50+ pre-configured APIs including Stripe, GitHub, Twilio, Resend and Clerk
  • Deterministic reproduction with shareable receipt URLs
  • Consumes no quota and needs no live credentials on the real provider

Installation

Set up this MCP server in your favorite AI agent — copy a ready-made configuration below.

Any MCP-compatible agent

Most agents (Claude, Cursor, Windsurf, VS Code, and more) read a standard mcpServers configuration.

  1. Open your agent's MCP configuration file.
  2. Merge the snippet below into it.
  3. Restart the agent — the "FetchSandbox MCP" tools will be available.
{
  "mcpServers": {
    "fetchsandbox-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "fetchsandbox-mcp@latest"
      ]
    }
  }
}

Description

FetchSandbox answers a problem that got sharper once agents started writing integration code: the code compiles, the happy path returns 200, and nobody finds out until production that the invoice.paid handler is not idempotent.

The server exposes an MCP interface over a simulated version of the APIs an integration talks to. An agent can import an OpenAPI spec, list the workflows the spec implies, and run them end to end against schema-validated responses — including the parts that are awkward to test for real: webhook delivery, retries, async state transitions, out-of-order events, duplicate deliveries and stale events. Because responses are deterministic, a bug reproduces the same way twice and can be shared as a "receipt URL" that documents exactly what happened.

More than fifty integrations ship pre-configured, among them Stripe, GitHub, OpenAI, Paddle, Twilio, Resend and Clerk. Nothing consumes quota on the real provider, and no live credentials are needed during development.

Setup

Published to npm as fetchsandbox-mcp and run with npx, so nothing needs installing globally. It works with Cursor, Claude Code, Claude Desktop, VS Code, Windsurf, Cline, Codex CLI and Continue.dev, and there is a web dashboard alongside it. Note that Cursor disables newly added MCP servers by default — enable it under Settings → MCP Servers or the tools will not appear.

Pricing

Free tier: 1,000 runs per month with all 50+ bundled APIs, CLI and MCP access, and OpenAPI imports. Pro is $5/month for 10,000 runs, with overage at $5 per additional 10,000, plus private integrations, webhook replay and run history. Enterprise adds unlimited runs, SSO/SAML, audit logs, an SLA and VPC or on-premises deployment. A "run" is one complete integration scenario, not a single API call.

Related MCP Servers

New

Replicate's official MCP server: search thousands of hosted models, read their schemas, and run predictions on image, video, audio and language models from inside an agent.

MCP: Hugging Face

by Hugging Face

New

Hugging Face's official remote MCP server — search and inspect models, datasets and Spaces, browse the Hub filesystem over hf:// URIs, and turn Gradio Spaces into callable tools.

MCP: Resend

by Resend

New

Resend's official MCP server for transactional and marketing email — send, schedule and batch messages, manage templates, contacts, broadcasts, automations, domains and suppressions from an agent.

New

Postman's official MCP server — reach your workspaces, collections, specs and environments from Claude Code, Cursor, VS Code or Copilot, in four tool tiers.

Browse all MCP servers →