Mastra

PASS consensus from 3 PASS.

The guessed create-project URL returned 404, but the docs root, agent page, model page, and develop page still provide one complete path. Get started gives a condensed agent-targeted install block plus a create mastra quickstart, and Develop gives the single mastra dev command that opens Studio on port 4111. One clear path: scaffold with npm create mastra@latest or create an Agent class manually, register it in the Mastra instance, and run Studio with npm run dev.

FAIL consensus from 1 PARTIAL, 2 FAIL.

The platform API says rate limits are unpublished, while the environments page says the environment allowance depends on the plan without stating a number. No page in the documentation states a request rate limit for the Mastra platform, and the only exact quotas (100K events, 24 CPU hours, 15 days retention on Starter) live on the marketing pricing page, which never states the per-plan environment count the docs defer to. The docs explicitly state that rate limits are not published contracts. No page documents exact rate limits or plan quotas for the Mastra platform. The Mastra Memory Gateway (a separate product at gateway.mastra.ai) has limits, but the main product docs have none.

PASS consensus from 3 PASS.

The error-handling page shows a failed fetch, step and workflow retries, result inspection, and onError handling. One page carries the whole correction path: failed result status, per-step error inspection, step-level retries, workflow-level retryConfig, onError, and a branch fallback, though there is no catalogue of error codes and retryConfig is absent from the Workflow class reference. Error handling is documented with retry policies at workflow and step level, lifecycle callbacks (onFinish, onError), conditional branching, and result status inspection. No standalone error-reference page exists, but the error-handling page serves that role for workflows.

PASS consensus from 2 PASS, 1 PARTIAL.

The JWT page covers package installation, token creation, server configuration, Bearer requests, and shared-secret verification. Install, configure, mint a token, and call with curl are all documented end to end, but no page states the signing algorithm or which claims are validated, so the token has to be produced through the jwt.io walkthrough rather than generated programmatically. Full path documented: install @mastra/auth, configure MastraJwtAuth with a shared secret, generate a JWT via jwt.io, send authenticated requests through MastraClient with an Authorization header. The same page covers server config and client usage.

PASS consensus from 3 PASS.

The guide installs @mastra/client-js, retrieves an agent, and calls generate('Hello'); the API reference supports the same string form. @mastra/client-js is named as the official client and the minimal getAgent().generate() example matches the string-prompt form shown in the Agents API reference and in the JWT guide. The official @mastra/client-js package is documented with a minimal working example. The API form (client.getAgent().generate()) is consistent across the MastraClient page, the Agents API page, and the server-side usage examples.