FriskAI launches with $3.6M to show enterprises what their AI agents are doing

FriskAI has launched with $3.6 million in pre-seed funding and a simple pitch: if enterprises are going to let AI agents loose in production, they should be able to see what those agents actually do.

That is the messy part of agentic software. Give the same agent different inputs, tools or objectives and it may take a path nobody planned. Traditional observability tools assumed software wouldn’t improvise. FriskAI is built around that gap.

The company’s software sits alongside an agent and records how it uses tools, including the arguments it sends, the responses it gets back and timing data. From that, it builds a behavioral profile for each agent by task and by tool, then compares new versions and new deployments against prior behavior. It also watches for changes in the scope of activity, the systems an agent reaches and how many calls it makes, with alerts sent in real time when something looks unusual.

FriskAI says it exposes the platform through Python and TypeScript SDKs, plus adapters for LangChain, Claude Agent SDK and Strands. The company is targeting healthcare, insurance and financial services, where buyers have to explain agent decisions to auditors. Sana Benefits is already using it, and staff software engineer Jason Moore said it has given his team the operational visibility they were missing and helped speed up troubleshooting.

The money came from MaC Venture Capital, with Wischoff Ventures, New Enterprise Associates partner Rick Yang and a group of angel investors joining in. The company’s website also lists Detroit Venture Partners. FriskAI, based in Los Angeles, is taking users through an early access program, and the new funding is going toward engineering, go-to-market hiring, customer deployments and more work on the behavioral analysis side.