Loading ToolBench
MCP server for AI-driven Cypress test execution. Run, debug, and iterate on E2E tests directly from your AI agent.
Cypress-mcp has 12 well-named tools with clear action verbs (discover, analyze, list, run, get, doctor). All tools have descriptions (avg ~120 chars, within baseline 34-392). Input schemas are present and typed for all tools. However, parameter descriptions are sparse or missing in several tools (e.g., cypress_run_spec lacks description for 'env' param; cypress_get_screenshot lacks description for 'includeData'). Output schemas are not documented in the source, LLMs cannot predict return structure. Error handling is minimal; no recovery guidance or actionable error messages visible. Tool composition is strong (discovery → run → inspect loop), but missing pagination for list tools and batch operations.
Analyze a single spec file and return its structure: describe blocks, test names, and line numbers. Useful for understanding test organization before running.
Discover all Cypress specs in the project and analyze their structure. Returns a list of spec files with test counts and describe blocks.
Run Cypress doctor to diagnose configuration and environment issues. Returns detailed diagnostics about the Cypress setup, dependencies, and potential problems.
Return the current Cypress MCP server configuration including project root, browser, timeouts, and directories.
Return environment variables and configuration available to Cypress tests. Useful for understanding test context and available secrets.
Get a compact debugging bundle for the most recent failure. Returns the failing test, error, stack hint, screenshot paths, relevant spec excerpt, and recommended next actions. This is the key tool for the fix-and-rerun loop.
Output schemas not documented. LLMs cannot predict return structure for any tool. Baseline requires 100% of A+ tools to have documented return types.
Parameter descriptions incomplete or missing. cypress_run_spec 'env' param has no description; cypress_get_screenshot 'includeData' lacks description. Baseline: 100% of A+ tool params have descriptions.
No error handling guidance. Tools return failures but do not tell LLM what to do next (retry, ask user, abort). Baseline pattern: recovery-guide.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | <=2025-11-25 | v2 |
Retrieve the full result object from the most recent test run, including all test details, failures, and metadata. Useful for detailed inspection after a run.
Retrieve screenshot files from the most recent test run or a specific spec. Returns file paths and optionally base64-encoded image data.
List all Cypress spec files in the project. Returns file paths and test counts.
Replay the exact last run_spec or run_test invocation. Removes friction when iterating on the same failure — no need to reconstruct arguments.
Run a single Cypress spec file headless. Returns structured results with pass/fail counts, failure details, screenshot paths, and recommended next actions.
Run a specific test by name within a spec file using grep. Returns structured results with recommended next actions for the agent.
List tools (cypress_list_specs, cypress_discover) lack pagination. No limit, offset, or cursor parameters. Returning unbounded lists risks context window exhaustion.
No batch operations. cypress_run_test runs one test at a time. Agents calling this in a loop waste tokens and latency vs a single batch_run_tests tool.