Playgent
ActiveSandboxes for AI agents
This company has previously operated under “Pharmie AI”. A rename frequently marks a pivot in positioning or product — useful raw material for variant ideas.
About
When enterprise agents fail in production, the conditions that caused the failure are critical: user state, data, tool calls, and context. Without a way to recreate these conditions, teams can't fix what broke. Playgent provides infrastructure to create high-fidelity sandbox environments for AI agents. Our environments come packaged as a single MCP url where tool calls are mocked, authentication is abstracted away, and test data can be initialized via natural language or JSON specifications. Teams can use Playgent to reproduce production failures in safe, scoped environments that accurately reconstruct state. Finally, export runs as OTel traces from Playgent's environment to plug directly into your observability tools or RL pipelines.
Founders · 2
Hi! I’m Aniruddh - co-founder of Playgent! We are building sandboxes for AI agents. I did 4 years of NLP research with publications in EMNLP/ACL and was on-route to pursue a PhD when I decided to join YC. In my free time, I like writing articles on Medium, playing basketball, and doing acrylic art. You can reach me at aniruddh@useplaygent.com!
Hey I'm Neer, cofounder of Playgent. We're building an Sandboxes for Agents! In college, I studied Math, Business, and CS at UT Austin and worked in product and engineering roles at multiple early stage AI startups. At Visa, I launched features used by thousands of banks and built internal tools. Feel free to reach out -- neer@useplaygent.com
Related startups
Real-world sandboxes to test agents and agent-facing software
sandbox infrastructure for AI agents
