FriskAI Secures $3.6M to Bring Observability and Audit Trails to Enterprise AI Agents
6 million in pre-seed funding, according to SiliconANGLE, pitching runtime intelligence for AI agents into a market that just minted Zenity's $125 million round two weeks earlier.
FriskAI exits stealth with $3.6 million in pre-seed funding, according to SiliconANGLE, pitching runtime intelligence for AI agents into a market that just minted Zenity's $125 million round two weeks earlier.
The Los Angeles startup, founded and led by CEO Neel Palrecha, sells observability software that runs alongside an agent and logs every tool call, argument, response, and timestamp. Each entry feeds a behavioral profile broken down by task and tool, and the platform flags when a new deployment starts drifting from the last one. Anomaly detection runs without preset rules, watching for shifts in scope, target systems, and call volume in real time.
What the product actually does
FriskAI ships Python and TypeScript SDKs with prebuilt adapters for LangChain, the Claude Agent SDK, and Strands. The pitch is deliberately unsexy: enterprises can already monitor servers, applications, and employees, but they still can't tell what an agent did at 3 a.m. or why it chose one tool path over another. The startup is targeting healthcare, insurance, and financial services, where audit trails aren't optional and procurement officers have questions. Health benefits provider Sana Benefits Inc. is an early user, with staff software engineer Jason Moore saying the platform delivered the operational visibility the team had been missing and cut troubleshooting time.
The cap table
MaC Venture Capital led the round. Wischoff Ventures joined, with NEA partner Rick Yang and a roster of angels filling out the syndicate. Detroit Venture Partners also appears on the company's website as a backer. Managing partner Marlon Nichols at MaC reached for the standard analogy bin, positioning FriskAI as the infrastructure layer for AI agents and comparing the opportunity to what AWS was to cloud and Cisco to the early internet. The pre-seed burns toward engineering hires, go-to-market, customer deployments, and further work on the behavioral analysis side. Access is currently gated through an early access program.
What to actually watch
The competitive picture is the real story. Zenity Ltd. closed $125 million on Aug. 3 to build a security layer for AI agents, and a string of smaller rounds has chased visibility into agentic workflows. Three-point-six million in pre-seed is a runway, not a moat. For enterprise buyers evaluating this category, the practical question is whether FriskAI's behavioral profiling layer is defensible against three converging threats: incumbents extending APM and observability suites into agent instrumentation, well-funded peers like Zenity with nine figures in the bank, and open-source instrumentation built on top of OpenTelemetry. At pre-seed, the company has a founder, a lead investor, and a demo. That buys time. It doesn't lock in a market.