# FetchSandbox

A testing service, driven from a coding agent through MCP or a CLI, that runs your third-party API integration (payments, webhooks and similar) against simulated provider 'twins', injects failures such as retries, duplicate or late requests, proposes fixes and returns a receipt of what was checked.

- Category: API docs and testing · by FetchSandbox
- Website: https://fetchsandbox.com/
- Live data: no
- Checked on the official pages: 2026-10-24

## API

- Official API: yes, docs https://fetchsandbox.com/mcp
- Key needed: yes
- Price: Free tier; limits: Free: 500 runs and 5 fixes a month per developer; Pro: 10,000 runs and 100 fixes a month
- Base URL: https://fetchsandbox.com/mcp/v1
- Formats: json

## Price and code

- Free tier: $0 per developer: 500 runs and 5 fixes a month, full catalog, MCP and CLI; paid from $200/month (https://fetchsandbox.com/pricing)

## Data and license

- License: FetchSandbox Terms (https://fetchsandbox.com/terms); covers Use of the testing service; your code and returned diffs/findings are yours; no reselling access, no live production credentials in sandboxes
- Commercial use OK
- Credit: not stated
- Embedding unclear
- Quote (api_official): "Free integration engine: connect through MCP and test supported integration workflows using controlled provider sandboxes." (https://fetchsandbox.com/llms.txt)
- Quote (api_key): "Local tools can load without a key, but verification calls require authentication." (https://fetchsandbox.com/mcp)
- Quote (free_tier): "Free $0 per developer 500 runs, 5 fixes a month Full catalog, MCP and CLI Bring your own agent" (https://fetchsandbox.com/pricing)
- Quote (paid_from): "Pro $200 per workspace, per month 10,000 runs, 100 fixes a month $3 a fix after; $5 per 10,000 runs after" (https://fetchsandbox.com/pricing)
- Quote (limits): "Free $0 per developer 500 runs, 5 fixes a month" (https://fetchsandbox.com/pricing)
- Quote (commercial): "Your code stays yours. You give us permission to process it only so far as we need to run the job you asked for — reading a repository, reporting findings, producing a diff. We do not train models on it." (https://fetchsandbox.com/terms)
- Quote (acceptable_use): "try to get around usage limits, or resell access as your own product upload credentials for a live production account into a sandbox" (https://fetchsandbox.com/terms)
- Quote (evidence_limits): "A pass applies only to the recorded version, observed scenario, assertions, and observation windows." (https://fetchsandbox.com/llms.txt)

## Rules for an agent

- Add the local MCP server (npm fetchsandbox-mcp, MIT client, run with npx) and set FETCHSANDBOX_API_KEY to the raw key (no 'Bearer' prefix); hosted clients use https://fetchsandbox.com/mcp/v1 with Bearer/API-key auth (choosing 'None' returns 401). Hosted tools cannot read your local files, so use the local server for code checks.
- Use test keys only; never upload live production credentials into a sandbox
- Treat a receipt as evidence for the recorded version and scenarios only, not general production approval
- Keep live billing disabled while testing against twins

## Agent access

- llms.txt: https://fetchsandbox.com/llms.txt
- MCP: Hosted https://fetchsandbox.com/mcp/v1 (Bearer/API key); local npm fetchsandbox-mcp via npx

Human page: https://agentgid.com/resources/fetchsandbox/
Source: AgentGid (agentgid.com). Facts are quoted from the site's own pages; check the license page before you ship.