Bespoke Labs raises $40 million to build training environments for AI agents

Bespoke Labs, a startup from Mountain View, California, raised $40 million to create environments for training and testing AI agents — the company announced this itself.

Why AI agents need training platforms

Modern AI agents can write code and answer questions, but according to Bespoke Labs observations, they still break down on long and complicated tasks that require many sequential steps. The company builds environments that train and test agents precisely on such scenarios — rather than on single questions, where modern models already perform well.

The idea is that evaluating an agent based on how it answers a single question is insufficient: a real work task can stretch over dozens of steps, include switching between tools and accumulating context, and it is on this path that agents most often lose track and make mistakes.

• Bespoke Labs startup is based in Mountain View, California
• Raised $40 million in total funding, according to the company's own statement
• The company was founded by Alex Dimakis and Mahesh Sathiamurthy
• The product — environments for training and testing AI agents on long multi-step tasks

Who is behind Bespoke Labs

The company was founded by Alex Dimakis and Mahesh Sathiamurthy. Bespoke Labs specializes in developer infrastructure for agents: instead of each company creating its own test scenarios to check agents for reliability, Bespoke Labs offers ready-made environments that reproduce long and "messy" work tasks on which today's agents most often stumble.

This approach saves time for teams implementing agents in production: instead of building their own test scenarios from scratch and manually marking where the agent succeeded and where it failed, they can rely on ready-made infrastructure for training and evaluation.

What this means

The Bespoke Labs round is part of a broader trend: as AI agents are being implemented in real work processes, demand is shifting from the models themselves to infrastructure for their verification — environments, benchmarks, and training platforms that show where an agent is truly reliable and where it breaks down over a long distance.

Frequently asked questions

How much did Bespoke Labs raise?

The startup raised $40 million in total, which the company announced itself.

Who founded Bespoke Labs?

The company was founded by Alex Dimakis and Mahesh Sathiamurthy.

What does Bespoke Labs do?

The company builds environments for training and testing AI agents on long, multi-step tasks, where modern agents most often lose track and make mistakes.

Why are investors investing in training environments rather than new models?

Because the models themselves already handle single requests fairly well, and the main barrier to implementing agents in real work is reliability on long multi-step scenarios, and it is precisely this that such environments test.