mastra

Run npm create mastra@latest for the interactive CLI, or install manually with npm install @mastra/core zod and the mastra dev dependency. Configure tsconfig.json with ES2022 modules and bundler resolution, then define agents in src/mastra and start with npm run dev.

Check embedded docs first in node_modules/@mastra/core/dist/docs, which match your installed version exactly. Use SOURCE_MAP.json to find type definitions, or fetch remote docs from https://mastra.ai/llms.txt when packages are not installed.

These errors usually mean the code uses an outdated API from older Mastra versions. Verify the current signature in the embedded docs under node_modules/@mastra/core/dist/docs and confirm all @mastra packages share the same version.

Mastra requires target ES2022, module ES2022, and moduleResolution bundler in tsconfig.json. CommonJS configurations cause import and module resolution failures, and package.json should include type module.

Memory requires a configured storage backend such as PostgresStore, a Memory instance assigned to the agent, and a consistent threadId passed to each generate call. Semantic recall additionally needs a vector store and embedder.

Start with the official migration guide at https://mastra.ai/llms.txt, update all @mastra packages together, run available codemods, then verify each changed API against the embedded docs in the updated node_modules before testing.