Bespoke Labs

Bespoke Labs builds realistic simulated business environments — codebases, microservices, logs, tickets, email and Slack — where enterprises and AI labs train and evaluate long-horizon agents before deploying them in production. It is aimed at AI teams and frontier labs that need agents to be dependable on real, economically meaningful workflows.

Bespoke Labs is a data-and-environments research lab, founded in 2024 by Mahesh Sathiamoorthy and Alex Dimakis and headquartered in Mountain View, California, that designs and builds reinforcement-learning environments which look and behave like real companies. Rather than chasing bigger models, Bespoke's thesis is that better training and evaluation environments produce more reliable agents: it constructs large codebases, microservices, realistic logs, tickets, email and Slack so autonomous systems can learn and be measured on long-horizon, real-world workflows before they reach production. The company is a core contributor to Terminal-Bench, a widely cited agentic-coding benchmark used by frontier labs, and built OpenThoughts, an open reasoning dataset downloaded hundreds of thousands of times. In July 2026 it announced $40M in Series A funding led by Wing VC (on top of an $8.25M seed led by 8VC), which it is using to expand its research team and scale its environment-building infrastructure for Fortune 500 enterprises and frontier labs.

Category

Agent Development

Pricing

Contact for pricing

Target Market

AI/ML Engineers, Data Scientists, CTOs, AI Research Labs

Founded

2024

Headquarters

Mountain View, California, USA

• ✓
  Realistic RL environments

  Simulated companies with codebases, microservices, logs, tickets, email and Slack that mirror real business settings.

• ✓
  Long-horizon agent training

  Environments designed for economically meaningful, multi-step workflows rather than toy tasks.

• ✓
  Agent evaluation and measurement

  Tools to test, measure and refine autonomous systems before production deployment.

• ✓
  Terminal-Bench contributions

  Core contributor to Terminal-Bench, a widely cited benchmark for agentic coding capability.

• ✓
  OpenThoughts dataset

  Open reasoning dataset downloaded hundreds of thousands of times and used across the AI research community.

• •
  Pre-production agent testing

  Run agents against company-like environments to catch failures before deploying them on live systems.

• •
  Post-training with reinforcement learning

  Use realistic environments to post-train models on long-horizon enterprise workflows.

• •
  Benchmarking agent reliability

  Measure and compare agent performance on standardized, real-world-style tasks.

Pricing is not publicly disclosed; Bespoke Labs works directly with enterprise and frontier-lab customers.

This page was written from 3 sources, 2 on domains other than bespokelabs.ai.

1. 1.bespokelabs.ai — bespoke labs raises 40m to build environments that enable revendor
2. 2.thesaasnews.com — bespoke labs raises 40m series a
3. 3.businesswire.com — Bespoke Labs Announces $40M to Build the Environments That T