Mastra
Founded 2024 · San Francisco, USA
A modern TypeScript framework for AI agents and applications, with an optional hosted cloud.
In brief
Mastra is an agent framework: A modern TypeScript framework for AI agents and applications, with an optional hosted cloud. Hosting is dual — self-host the open-source typescript library on your own infra; a free tier is available to start; and it's native on MCP. Best suited to teams that want to build agent logic in their own codebase.
What it is
Mastra is an open-source (Apache-2.0 core) TypeScript framework for building AI agents and applications — agents, workflows, memory, and observability — with an optional managed Mastra Cloud for deploying and observing them.
At a glance
7/7 facts sourced- Pricing
- Self-hosted is free (Apache-2.0). Cloud Starter $0/mo with usage metering, Cloud Teams $250/mo, Enterprise custom; add-ons include memory at $10/1M tokens.source ↗
- Hosting model
- Dual — self-host the open-source TypeScript library on your own infra, or use managed Mastra Cloud.source ↗
- Memory & state
- Message history, working memory (persistent structured user data), semantic recall (embedding-based retrieval), and observational memory; requires a storage provider (PostgreSQL, LibSQL, or Redis).source ↗
- MCP support
- Nativesource ↗
- Integrations
- Tools and MCP servers, plus framework adapters for Next.js, React, Astro, Express, SvelteKit, and Hono, and data-source connectors.source ↗
- Model providers
- A model router to 90+ providers via a single provider/model API — OpenAI, Anthropic, Google, DeepSeek, Groq, Mistral, xAI, plus gateways like OpenRouter and any Vercel AI SDK model.source ↗
- Open source
- Yessource ↗
Editorial scorecard
Overall 4.2 / 5
Our opinion — the same published 0–5 rubric is applied to every platform. See the rubric.
Strengths
- Core is open source (Apache-2.0) and free to self-host.
- Native MCP plus a model router across 90+ providers with fallback.
- Bundles several memory types — working memory, semantic recall, and observational memory.
Trade-offs
- TypeScript-only, which excludes teams standardized on Python.
- Memory requires an external storage backend (PostgreSQL, LibSQL, or Redis).
- Mastra Cloud uses usage-based metering, making cost harder to predict than a flat fee.
Our take
Editorial opinion — not a hands-on test resultIn our opinion, Mastra is one of the more polished options for TypeScript teams: the developer experience and MCP support are strengths, and the model router with fallback is genuinely useful. The trade-offs are language lock-in to TypeScript and the need to bring your own storage backend for memory. Its integration catalogue is narrower than the largest frameworks, so check that your specific connectors exist.
FAQ
- Is Mastra free and open source?
- The core is Apache-2.0 and free to self-host; enterprise directories are source-available, and Mastra Cloud has paid tiers.
- Which LLM providers does Mastra support?
- 90+ providers via its model router, plus any model available through the Vercel AI SDK.
Compare Mastra
Sources
- Mastra homepage · accessed 2026-07-01
- Mastra pricing · accessed 2026-07-01
- Mastra memory docs · accessed 2026-07-01
- Mastra on GitHub (Apache-2.0 core) · accessed 2026-07-01
The Agent Examiner editorial · last updated 2026-07-01. Facts are drawn from each vendor’s own public materials; scores are our editorial opinion (methodology).
Product and company names are trademarks of their respective owners and are used here only to identify the products under discussion.