CLI reference
FetchSandbox CLI
Run any FetchSandbox workflow from your shell. Pipe results into CI/CD. Drop the verified-skill file into Cursor or Claude Code so the agent runs the same commands you do.
Install
Quickstart
Commands
17 commands total. The most-used is run — it has the most flags and is what your agent or CI invokes.
Create a sandbox from an OpenAPI spec (URL or file path).
List your sandboxes.
Delete a sandbox.
-f, --forceList the endpoints in a sandbox.
-s, --search <term>Print the docs for a sandbox to your terminal.
List workflows for a sandbox.
-n, --name <name>Send a raw HTTP request through the sandbox.
-d, --data <json>Execute a workflow. The most-used command — supports failure scenarios, JSON output for CI/CD, and proof-checking.
--all--scenario <name>--verbose--json--preflightPrint a one-shot status snapshot of the sandbox.
Show recent request logs.
-f, --follow--status <code>-l, --limit <n>View or set the active failure scenario for a sandbox.
-n, --name <name>Inspect resource state inside the sandbox.
--id <resource-id>Reset a sandbox to its initial state.
Show recent mutations applied to the sandbox.
--last <n>Tail webhook events as they fire.
--verbose--event-type <type>Generate a verified-skill file for an AI agent (Cursor / Claude Code).
--output <file>Health check — verify the sandbox is reachable and healthy.
Common patterns
CI/CD: assert your integration before deploy
Run the workflow with --preflight --json to get a pass/fail you can pipe into your pipeline.
Test failure handling without touching production
Switch to a failure scenario, run the workflow, see how your handler behaves.
Agent integration via verified-skill files
Generate a skill file, drop into your AI agent's context. The agent reads the spec, calls the sandbox, integrates without inventing endpoints.