AI Potluck
Contribute
Back to Gap Map Product / UX / Orchestration & agents

Mastra

Mastra
open core / Overall score: 4.0(strong)

TypeScript-native framework for building agents and agentic applications. It covers agents with tools and instructions, graph-based workflows with explicit control flow, conversation memory with semantic recall, model routing across many providers, and built-in evaluations, tracing and human-in-the-loop. It runs on Node.js, integrates with React and Next.js, or deploys as a standalone server.

GitHub's license classifier cannot read the split license file; the score was read from the file body.

Openness

4 high confidence
4.0
license
Apache-2.0(OSI, everything outside `ee/`)
source
public
core features withheld
yes — RBAC, SSO, FGA and the agent builder live in ee/-scoped directories under the Mastra Enterprise License, and the auth features are switched on at runtime by MASTRA_LICENSE_KEY, validated against the Mastra license server

Mastra's framework is Apache-2.0 outside a set of ee/ directories that carry RBAC, SSO, fine-grained authorization and the agent builder under the separate Mastra Enterprise License, with the auth features activated only once MASTRA_LICENSE_KEY validates against Mastra's license server. Because those are real features withheld from the open license, not merely a hosted platform sold beside it, Mastra is open core rather than fully open source.

  • https://github.com/mastra-ai/mastra/blob/main/LICENSE.md recorded 2026-07-30

    LICENSE.md: "All content that resides under any directory named 'ee/' within this repository" - listing packages/core/src/auth/ee/ and packages/server/src/server/auth/ee/ - "is licensed under the license defined in ee/LICENSE", while everything "not mentioned" in those directories or restrictions "is available under the 'Apache License 2.0'". This preamble ahead of the Apache text is why GitHub's classifier reports NOASSERTION for the repo.

  • https://mastra.ai/ recorded 2026-09-29

    The site FAQ says "core framework is open source under the Apache 2.0 license. Enterprise features are source-available under the Mastra Enterprise License".

  • https://raw.githubusercontent.com/mastra-ai/mastra/main/ee/LICENSE recorded 2026-09-29

    The ee/LICENSE states "Enterprise Features are made available in source-available form. They are not open source." and bars Production Use without a license.

  • https://raw.githubusercontent.com/mastra-ai/mastra/main/LICENSE.md recorded 2026-09-29

    The root LICENSE.md maps the repository to two licenses: "All content that resides under any directory named" ee/ directories "is licensed under the license defined in" ee/LICENSE, while "Content outside of the above-mentioned directories or restrictions is" under the Apache License 2.0.

  • https://raw.githubusercontent.com/mastra-ai/mastra/main/packages/_internals/auth/src/ee/license.ts recorded 2026-09-29

    The license.ts source opens "License validation for EE features." and treats "Server says invalid/revoked/expired" as EE features disabled, so a key checked against the license server switches the features on.

  • https://raw.githubusercontent.com/mastra-ai/mastra/main/README.md recorded 2026-09-29

    The README states "This repository uses a dual-license model:" Apache-2.0 for the core framework and the Mastra Enterprise License for ee/ directories that "require a valid enterprise license for production use".

  • https://registry.npmjs.org/@mastra/core recorded 2026-09-29

    Registry metadata for @mastra/core reads "bugs":{"url":"https://github.com/mastra-ai/mastra/issues"},"license":"Apache-2.0","homepage":"https://mastra.ai", so the published package carries the OSI license.

Adoption

4 high confidence
4.0

Adoption is measured by npm downloads of @mastra/core, the package everything else in the framework resolves to; the separate mastra package pulls in @mastra/core transitively, so summing the two would double-count the same installs. Raw download volume also includes CI and mirror traffic, so this reading measures install volume rather than a count of individual users.

Capability

4 medium confidence
4.0

Mastra is level with LangChain and Pydantic-AI on feature coverage, and a tier below LangGraph and LlamaIndex, which have deeper ecosystems and longer track records. No first-party SWE-bench or GAIA result exists for the framework, so the score reads from features rather than a measurement.

  • https://mastra.ai/ recorded 2026-09-29

    The site describes "Mastra workflows also support human-in-the-loop functionality", "model routing connects to 90+ providers through one standard interface" and memory with "semantic recall"; no benchmark figure is published on the page.

Verified 2026-09-26