← All organizations

TypeScript AI agent framework and platform

Mastra

Mastra builds an open-source core framework in TypeScript and an accompanying platform for developers creating AI agents, customer support tools, internal assistants, and workflow automation. It combines typed agents, tools, memory, and workflows with evaluation and tracing. Mastra Studio provides cloud-based or self-hosted evaluation and observability, Mastra Server deploys agents and workflows, and Memory Gateway supplies agent memory across frameworks. Developers can integrate Mastra into React, Next.js, and Node applications or deploy it as a standalone server.

Founded in 2024 by Sam Bhagwat, Abhi Aiyer, and Shane Thomas, Mastra grew out of the team's work on Gatsby. Bhagwat is CEO, Aiyer is CTO, and Thomas is Chief Product Officer. Its Observational Memory system uses background Observer and Reflector agents to compress conversation history into a structured observation log. This maintains context that can be cached across turns without retrieving and injecting memories for each new prompt.

In 2026, the company reported that Marsh McLennan's Mastra-built enterprise search served more than 100,000 people daily, illustrating the scale of one customer deployment. Mastra raised a $22 million Series A led by Spark Capital in 2026, bringing total funding to $35 million.

mastra.ai

2 talks

Newest first

1 speaker at AIE

Affiliations reflect their AIE appearances, not necessarily current employment.

Start here

  1. Agents vs Workflows: Why Not Both?

    Start here to understand why autonomous agents and structured workflows can be composed together, including supervisors coordinating specialized agents.

    Sam BhagwatAI Engineer World's Fair 2025

  2. Every Harness Will Become A Claw

    Continue here for a breakdown of harness capabilities such as planning, parallel subagents, and skills, and Bhagwat’s proposed progression toward personal agents.

    Sam BhagwatAI Engineer World's Fair 2026

Messages from the stage

Choose primitives around execution needs

Bhagwat distinguishes iterative tool-using agents from dependency-ordered pipelines. His discussion also questions cumbersome graph-oriented developer APIs, drawing on lessons from Gatsby and GraphQL.

Extend harnesses beyond coding sessions

Bhagwat describes cloud sandboxes, Slack collaboration, and messaging channels as infrastructure for continuously available agents. He discusses OpenClaw and Hermes Agent as examples and predicts consolidation across the agent ecosystem.

Affiliations reflect each recorded session, not necessarily current employment.

Company sources · checked 2026-08-27