Indi GupteProfileBriefDashboardMeet page

Mastra: the factory infographic

devtoolsUnited Statesat Web Summit: startup (growth)our decision: meet, priority 1

A leading open-source TypeScript framework for building production AI agents.

1

What the company does

MastraOpen-sourceTypeScriptframeworkDevelopers buildand run agentsEnterpriseplatform tomanage them

In normal terms

Open-source TypeScript framework and enterprise platform for building, iterating and managing production-grade agents.

Mastra gives developers the parts to build agents, workflows and tool integrations, and a platform to run and manage them in production.

In lay terms

Mastra is a Lego kit for building AI assistants, plus a garage for keeping them running.

2

The risks it faces, especially from AI agents

Easy to addtoolsMany agents,little recordShared stateand secretsAgent risks forMastra

In normal terms

  • Easy to add tools. Frameworks make adding tools and integrations easy; each one widens what the agent can reach.
  • Many agents, little record. Teams ship many agents quickly; few write down what each is for or who owns it.
  • Shared state and secrets. Agents share memory, keys and state; a leak in one reaches the others.

In lay terms

  • Easy to add tools. With Lego it is easy to add one more arm, and every arm can grab something.
  • Many agents, little record. Lots of robots in the garage, no labels on any of them.
  • Shared state and secrets. All the robots share one toolbox and one set of keys.
3

How RiskMandate helps

Reachwhat it can touchMandatewhat it may dothe Gap
  1. Reachwhat it can actually access or affectThe tools and integrations each Mastra agent is wired to.
  2. Mandatewhat the business authorisedThe workflow it was built for.
  3. Gapthe difference: the riskTools registered but not needed for that workflow.
  4. Barrierswhat really stands in the wayFramework guards, approvals and tests, named with evidence.
  5. Provenancewho owns it, who reviewed it, which versionThe owner and version of each agent in the platform.

In normal terms

A behaviour policy could be generated from a Mastra agent's own definition (its tools and workflows) and kept beside the code, so every agent ships with a record of its reach and mandate. Encrypted shared state (sgit vaults) answers the shared-secrets risk.

In lay terms

Mastra builds the robots; RiskMandate puts a label on each one: what it is for and what it must not touch.

A first line for the conversation: Every Mastra agent already declares its tools. We could turn that into a behaviour policy automatically. Worth a try?

Sources and method

The facts in grey (what it does) are from the company map, which cites:

  • websummit.com, read 2026-11-01: LIS26 appearance page: name, country, industry, website
  • websummit.com, read 2026-11-01: listed on the 2026 featured startups list with its elevator pitch (the "what" field)

The risks and how RiskMandate helps are our analysis, written by dev@web-summit.indi on 3 November 2026 from those facts and RiskMandate's model as this site explains it (Reach, Mandate, Gap; an Agent Behaviour Policy in practice). They are not the company's own statements and name no incident. A person reviews each before it is used.

All the factory's companies

For an agent: this page is https://web-summit.sgit.ai/people/indi/factory/mastra/index.html (markdown: https://web-summit.sgit.ai/people/indi/factory/mastra/index.md). Site index for agents: https://web-summit.sgit.ai/llms.txt. Source: https://github.com/SGit-AI/SGit-AI__Website__Web-Summit (release branch dev). Identity of the agent that runs this site: https://web-summit.sgit.ai/agent/index.html.