FriskAI, a Los Angeles company that monitors autonomous AI agents as they operate, has closed a funding round led by MaC Venture Capital. Wischoff Ventures and Detroit Venture Partners joined the financing. The startup declined to disclose its valuation.
The company's platform sits around agent runtimes, capturing detailed traces of each action an autonomous system takes in production: which tools it calls, what arguments it passes, what it returns, how long operations take. From those traces, FriskAI builds behavioral profiles and flags when an agent deviates from expected patterns. Teams don't set thresholds or write detection rules themselves; the system learns normal behavior and surfaces anomalies automatically.
FriskAI offers software development kits for Python and TypeScript, with native adapters for LangChain, Claude Agent SDK, and Strands. According to the company's website, the monitoring layer adds minimal performance overhead to tool calls.
The platform watches for what FriskAI calls Scope Deviation, Operation Deviation, and System Deviation, along with anomalies in argument patterns and call volumes. Traces feed into dashboards that break down tasks and tools by agent, detect changes across versions and deployments, and send alerts when behavior shifts unexpectedly. The company says enterprises can plug monitoring into any agent architecture with minimal setup.
Neel Palrecha founded and leads FriskAI. Since 2019 he has served as chief technology officer at PlayVS. Before that, Palrecha held engineering leadership roles at Snapdocs and Headspace, and earlier worked at Apple and Beats Music. He also founded an earlier venture called Chirply. FriskAI was founded in 2025 and currently employs between 2 and 10 people, according to its LinkedIn profile. Jonathan Ikpeazu has maintained multiple releases of the company's Python SDK on PyPI.
The fundraise arrives as a wave of observability startups targets the growing enterprise adoption of autonomous agents. Laminar disclosed a $3 million round in March, positioning its product for long-running agents. Respan announced $5 million that same month from Gradient, Y Combinator, and others for what it described as proactive AI observability. Definity raised $12 million in April for an agentic data engineering platform that includes runtime intelligence features.
Established security vendors have moved quickly to add agent monitoring capabilities. CrowdStrike said in March it extended AI agent discovery and runtime threat detection across endpoints, SaaS environments, browsers, and cloud infrastructure. Aqua Security announced the following month that customers could build agents interacting with its runtime intelligence and enforcement systems.
According to a paper titled "Runtime Compliance Verification for AI Agents" that appeared on arXiv, offline testing methods fall short for AI agents, and runtime verification becomes necessary once agents operate autonomously.
FriskAI is running an early-access program for its platform. The company has not yet disclosed customer names or detailed growth plans.
