Guides building Mastra agents, workflows, tools, and memory with verified current APIs.
npx skills add https://github.com/mmnavarr/harness --skill mastra-mmnavarr
Please help me install this Agent Skill. Skill: mastra Source: https://github.com/mmnavarr/harness/tree/main/skills/mastra Command: npx skills add https://github.com/mmnavarr/harness --skill mastra-mmnavarr
node_modules/@mastra/*/dist/docs/ matching the installed version, with remote docs from mastra.ai as fallback. - Project scaffolding and migration: Step-by-step project creation with correct TypeScript ES2022 config, plus version upgrade workflows using official migration guides and codemods. - Error diagnosis and model validation: Catalogs common errors (tool registration, memory persistence, workflow commits, storage connections) and verifies provider/model names via a registry script. - Use Case: A developer asks to add memory to a Mastra agent. The Skill checks installed packages, reads the embedded type definitions for the current Memory API, and writes code that compiles against the actual installed version. ## Quick Start Ask the AI to create a Mastra agent with a weather tool and verify the model name against the provider registry before writing any code.High-intent search queries and answers about installing and using this skill.
Run npm create mastra@latest for the interactive CLI, or set up manually by installing mastra and @mastra/core, configuring tsconfig.json with ES2022 modules, and creating an agent, tool, and Mastra entry point. Then run npm run dev to open Studio at localhost:4111.
Use an agent for open-ended tasks that require decisions and tool usage, such as support or research assistants. Use a workflow for defined multi-step processes like pipelines, approvals, and ETL where steps execute in a structured sequence.
Mastra's model router requires the "provider/model-name" format, such as "openai/gpt-5.4" or "anthropic/claude-sonnet-4-5". Run the provider-registry script to verify provider keys and model names, since available models change frequently.
Memory requires a configured storage backend, a Memory instance assigned to the agent, and a consistent threadId passed to each generate call. Without storage or a stable threadId, conversation history is lost between calls.
These errors usually mean your code uses an outdated API, since Mastra changes rapidly between versions. Check the embedded docs in node_modules/@mastra/core/dist/docs/ for the installed version's actual signatures rather than relying on memory.
Start with the official migration guide at mastra.ai/llms.txt, update all @mastra packages together, run any available codemod, then verify new API signatures against embedded docs. Test with npx tsc --noEmit and npm run dev afterward.
langchain-aiScaffolds a minimal local LangGraph agent in TypeScript following the official quickstart.
Celia-12Add standardized log statements to Mastra tools, agents, and workflows.
diShine-digital-agencyStore and recall semantic memories with metadata using Qdrant MCP.
langwatchInstrument Python and TypeScript codebases with LangWatch tracing and observability.