Arga Labs, a San Francisco-based AI infrastructure startup, has raised $10 million in seed funding to help companies train and test autonomous AI agents in safe, simulated environments.

The round was led by General Catalyst, with participation from BoxGroup, Emergence Capital, Gradient Ventures, and SV Angel.

Arga Labs develops sandbox environments that create digital replicas of enterprise software and APIs, including tools such as Salesforce, Workday, Stripe, Slack, and corporate email systems. These replicas allow AI agents to interact with realistic software environments without affecting real company data, payments, or communications.

The platform helps engineering teams train, stress-test and evaluate AI agents across complex workflows. It can identify problems such as incorrect tool calls, operational errors and security vulnerabilities before AI agents are deployed in real-world environments.

Arga’s simulated environments also allow developers to run large numbers of AI training interactions without facing the API rate limits of live software systems.

Founded in 2025 by CEO Phillip Li and CTO Akira Tong, Arga Labs will use the new funding to expand its engineering team, improve its software simulation technology, and accelerate product adoption among enterprise engineering teams.