Control plane that discovers, simulates, and enforces boundaries on autonomous AI agents

CompFly AI is a San Francisco company founded in 2026 to address a control problem that arrives with agentic systems: an agent that can call tools and act on data is functionally an employee with credentials, but without a manager, a job description, or a record of what it did. Co-founder Anand Salodkar came to the problem from fifteen years in compliance, audit, and cybersecurity inside regulated financial services, including roles at Ernst and Young, Franklin Templeton, and Dolby, and the product reflects that background: the questions it answers are the ones an auditor asks.

The platform connects to more than 50 systems spanning cloud providers, model APIs from OpenAI, Anthropic, and Google, agent frameworks such as LangChain and LlamaIndex, and enterprise applications including Salesforce and ServiceNow. Discovery inventories the agents actually running in the estate. Runtime governance evaluates each intended action against tool and data boundaries and either permits it, blocks it, or routes it for human approval before execution rather than after. Observability traces the full path of an agent's work, and a simulation environment lets teams probe agent behavior adversarially before deployment. Decisions are written to a TrustLedger for audit. The company is SOC 2 Type II certified and a member of the Coalition for Secure AI.