Is Mastra open source? License, funding, and the GitHub surge

Mastra, an open-source TypeScript framework for agentic applications from builders with a long track record in JavaScript tooling, drew intense GitHub attention in 2026 alongside heavy weekly npm traction and a stable 1.0 launch discussed on Hacker News. The attention lands at a moment when teams want fewer one-off glue scripts and clearer runtime primitives around routing, workflows, eval hooks, and observability exports. (Sources: Wellows AI startups roundup, Show HN Mastra 1.0)

The important idea is not “another trending repo badge.” It is consolidation pressure: when agents graduate from demos to scheduled jobs and customer-facing surfaces, integration beats a pile of half-connected SDK snippets. Mastra’s backers emphasize broad distribution and an Apache-licensed core as part of that story. (Sources: Mastra About page, Show HN Mastra 1.0)

The reporting spine below blends one independent startup roundup, Mastra’s own company narrative, maintainers’ Show HN notes for Mastra 1.0, and the public repository page. (Sources: Wellows AI startups roundup, Mastra About page, Show HN Mastra 1.0)

Is Mastra open source?

Yes, with one boundary worth knowing before you deploy. The mastra-ai/mastra repository states that "The core framework and the vast majority of this codebase is open source under Apache-2.0." Code under directories named ee/ is treated differently: it is source-available under the Mastra Enterprise License, and those features "require a valid enterprise license for production use but can be freely used for development and testing." (Source: Mastra GitHub)

In practice you can build, ship, and commercially operate on the Apache-2.0 core without paying or asking permission, which is the answer most teams are looking for. What to check before a production rollout is whether any feature you depend on lives under ee/, because that is where the license changes. Mastra's own company page describes the project as "the leading open-source TypeScript agent framework." (Sources: Mastra GitHub, Mastra About page)

Operator note (first-hand): on a repository check dated 2026-07-27, the public mastra-ai/mastra page rendered 26.6k stars and showed the dual Apache-2.0 plus Mastra Enterprise License arrangement in its license summary. That is up from the roughly 23.7k stars recorded in the May 2026 spot check noted below, which is a useful reminder that star snapshots age fast. (Source: Mastra GitHub)

What shipped

Startup trackers flagged Mastra when its repository reportedly jumped from 1,500 to 7,500 GitHub stars within a single week during a viral window, pointing readers toward discussion threads and the canonical repo listing. (Source: Wellows AI startups roundup)

On Mastra’s site, the company describes itself as a 2024-started TypeScript agent framework team with millions of downloads each month. On a 2026-07-27 check that page listed 34 people working remotely and $35m raised from backers including Spark Capital, Gradient, basecase, and Y Combinator, both figures higher than when this story first published. It also leans on prior open-source credibility by naming the team behind Gatsby. (Source: Mastra About page)

In the Show HN: Mastra 1.0 thread, maintainers said Mastra had reached 300k+ weekly npm downloads and 19.4k GitHub stars at the time of that post, ships under Apache 2.0, and named production mentions including Replit, PayPal, and Sanity. The 1.0 announcement stresses native model routing described as 600+ models across 40+ providers via a model string with TypeScript autocomplete and fallbacks, plus guardrails, scorers, tracing exporters, memory processors, a .network() routing primitive, and server adapters for Express and Hono. Getting started is documented as npm create mastra@latest. (Source: Show HN Mastra 1.0)

The same ecosystem summaries note Mastra integrates with React, Next.js, and Node.js, can run as a standalone server, and supports MCP servers via the Model Context Protocol. (Source: Wellows AI startups roundup)

The public GitHub repo bills Mastra as a framework for AI-powered applications and agents on a modern TypeScript stack, explicitly tying the project to the team behind Gatsby. (Source: Mastra GitHub)

Why GitHub noise and npm signal both matter for Mastra right now

Stars move fast and incentives around them are uneven. They still act as a coarse discovery funnel when practitioners compare dozens of similarly named “agent” repositories across languages and runtimes. Sudden star velocity can signal when a project crosses from insider chatter into broader developer feeds, even though stars are not operational proof on their own. (Source: Wellows AI startups roundup)

Weekly package downloads track repeated installs and upgrades across CI and laptops at a cadence closer to real adoption attempts than a one-time star click. Maintainer figures citing 300k+ weekly npm downloads alongside 1.0 stable framing give teams a second axis to cross-check enthusiasm against registry pulls. (Source: Show HN Mastra 1.0)

Why this matters:

  • Discovery versus validation: star spikes show where attention moved; registry cadence hints whether engineers came back after a first experiment failed. (Sources: Wellows AI startups roundup, Show HN Mastra 1.0)
  • Single metrics lie: no one number greenlights production; combine maintainer claims, independent summaries, and your own smoke tests before you bet a roadmap on them. (Inference: editorial judgment)

Operator note (first-hand): On a spot check of the public mastra-ai/mastra repository page in May 2026, the banner showed roughly 23.7k stars in the rendered summary line for that page load. Star counters move continuously, so treat any snapshot as a hint, not a benchmark. (Source: Mastra GitHub)

Practitioner payoff: workflows, routing, and guardrails in one TypeScript-shaped stack

Teams adopting agent patterns quickly collide with the middle layer that demos skip: workflow boundaries, tool boundaries, and observability boundaries. Mastra’s 1.0 messaging emphasizes primitives aimed there: multi-agent workflows, eval surfaces tied to agents, local studio inspection, and tracing hooks aimed at third-party observability tools. (Source: Show HN Mastra 1.0)

Practitioner payoff:

  • Model routing ergonomics: maintainer copy stresses large model catalogs through a single string API with autocomplete and fallbacks, which matters when providers rename models weekly and reroutes become routine. (Source: Show HN Mastra 1.0)
  • Operational guardrails: low-latency input and output processors for prompt injection detection, PII handling, and moderation appear as first-class themes in the launch narrative, reflecting production anxieties rather than demo gloss. (Source: Show HN Mastra 1.0)
  • Routing agents: a .network() primitive is positioned as a runtime hierarchy decision maker across workflows and agents when teams want dynamic delegation rather than static graphs only. (Source: Show HN Mastra 1.0)

Company positioning underscores distribution scale (“millions of downloads each month”) and pedigree (“team behind Gatsby”) as shorthand that these maintainers shipped large OSS surfaces before the current agent hype cycle. That does not guarantee fit for your architecture, but it explains why enterprise-facing narratives cite names like PayPal alongside smaller shops. (Sources: Mastra About page, Show HN Mastra 1.0)

Ecosystem positioning: frameworks, SDKs, and MCP-shaped glue

Mastra’s lane is TypeScript-first application framing: workflows, studio, storage-backed behaviors, and packaged operator ergonomics. That sits next to, not on top of, thinner client libraries that stop at streaming tokens and tool calls. Maintainer replies in the Show HN thread acknowledge proximity to Vercel AI SDK evolution while arguing Mastra’s integrated storage, studio, evals, and workflow suspend or resume paths reflect different architectural centers. (Source: Show HN Mastra 1.0)

Independent summaries tie Mastra to MCP servers. Protocol-first tooling shifts trust boundaries toward tool hosts and transport choices. If you stack Mastra next to protocol-heavy setups, STDIO versus hosted server tradeoffs belong in threat modeling, because the easiest wiring path is rarely the safest default for privileged tools. (Sources: Wellows AI startups roundup, Show HN Mastra 1.0)

Decision rule for teams:

  • If your organization already centralizes agent governance around graph workflows and explicit tool contracts, compare Mastra’s workflow story against Microsoft’s agent-framework trajectory as a checklist of primitives, not vendor logos. Past AgenticWire coverage on graph workflows and MCP offers a parallel baseline. (Sources: Show HN Mastra 1.0, Inference: editorial comparison)
  • If stakeholder conversations hinge on harness versus sandbox language for long-lived agents, align vocabulary with how OpenAI’s Agents SDK story framed harness and sandbox responsibilities so everyone shares one diagram. (Inference: editorial guidance)
  • If MCP is how your enterprise expects tools to surface, read Mastra’s positioning next to practical MCP adoption guidance and STDIO risk reporting so “supports MCP” does not get mistaken for uniform safety. (Sources: Wellows AI startups roundup, Inference: editorial guidance)
Consolidation wins when teams can name three boundaries (data, tools, observability) without inventing a private taxonomy every sprint.

Context

The Mastra narrative fits a recurring 2026 pattern: vendors and OSS projects racing to package agent primitives teams can adopt without standing up entirely bespoke platforms. Mastra’s maintainers emphasize TypeScript ubiquity and batteries-included posture; public threads also surface fair warnings about opinionated stacks and escape hatches when patterns do not fit. Those tensions are normal for fast-moving framework waves. (Source: Show HN Mastra 1.0)

Readers weighing MCP governance and harness semantics alongside Mastra often read Microsoft’s graph workflow and MCP angle (Microsoft Agent Framework 1.0 workflows and MCP), harness versus sandbox framing for long-running agents (OpenAI Agents SDK harness versus sandbox), STDIO configuration risk when tool wiring behaves like command execution (MCP STDIO configuration risk), and practical MCP adoption momentum (MCP adoption accelerates practical guides). Those pieces situate Mastra next to broader platform moves without treating any vendor snapshot as interchangeable. (Inference: editorial reading map)

Adoption notes

Pilot Mastra the way you would pilot any ops-touching framework: pin versions, mirror registry installs in CI, and replay tracing exports into your existing observability vendor rather than treating studio defaults as production truth. Maintainer messaging highlights tracing exporters toward tools like Langfuse and Braintrust, which implies your existing sinks matter as much as framework features. (Source: Show HN Mastra 1.0)

Treat star-driven urgency as a scheduling signal, not a procurement mandate. Combine Show HN figures with npm cadence and your own smoke tests (npm create mastra@latest is the documented entry point in the maintainer thread) before you promise timelines to application teams. (Source: Show HN Mastra 1.0)

When investors or executives ask about funding claims, stick to what listings and Mastra’s own narrative actually say rather than extrapolating beyond those sources. (Sources: Wellows AI startups roundup, Mastra About page)

FAQ

Is mastra AI free?

The framework is free to use. The core is Apache-2.0 licensed, so there is no fee to run, modify, or ship it commercially. Features under ee/ directories are the exception: they need a paid enterprise license for production use, while staying free for development and testing. (Source: Mastra GitHub)

Is mastra production ready?

Mastra shipped a stable 1.0, and maintainers named Replit, PayPal, and Sanity among production mentions in the Show HN launch thread. Treat named logos as a starting signal rather than proof of fit: pin versions, mirror registry installs in CI, and run your own smoke tests before committing. (Source: Show HN Mastra 1.0)

Who is the CEO of Mastra?

Sam Bhagwat is founder and CEO. Abhi Aiyer is CTO and Shane Thomas is co-founder and chief product officer. The same three previously built Gatsby, which is the open-source track record the company leans on when it positions Mastra to enterprise buyers. (Source: Mastra About page)

Related coverage

References