Archestra.AI raises $10M seed led by 20VC
What's the deal? London-based Archestra.AI has raised $10M in a seed round led by Harry Stebbings' 20VC. The round also includes 20 Product, Visible VenturesDealroom has a profile for this one. Try Dealroom →, and Tenacity CapitalDealroom has a profile for this one. Try Dealroom →, alongside angels such as Datadog chief executive Olivier Pomel and HubSpot chief marketing officer Kieran Flanagan.
The startup, founded in 2025 by Grafana Labs alumni, has raised $13.5M to date. It will use the fresh capital to grow its go-to-market and engineering teams and accelerate enterprise deployments.
Archestra.AI builds an open-source platform that lets enterprises securely connect sensitive data to AI agents. The idea: give companies guardrails so employees can spin up agents that work with legal, HR, and other confidential data — and even communicate with external parties — without risking leaks.
Why now? Enterprises are deploying AI agents for basic tasks but remain wary of letting them touch sensitive information. Fears of hallucinations, rogue behaviour, and data exposure have kept most companies from scaling agentic AI across business-critical functions.
"Fortune 500 companies don't feel comfortable relying on LLM companies like Anthropic and OpenAI solely to ensure AI agents are scaled safely," said chief executive and founder Matvey Kukuy. He added that enterprises want solutions that are independent of any single model provider, avoiding vendor lock-in.
What could go wrong? The startup is a seven-person team competing in one of the hottest — and most crowded — spaces in enterprise software. Proving that its guardrails truly prevent data leaks at scale will be the ultimate test, especially as larger incumbents rush to add similar security layers to their own agent platforms.
The signal: With four Fortune 500 clients already signed, Archestra.AI's early traction suggests enterprises are willing to bet on specialist, model-agnostic tooling rather than rely on AI providers' own safety promises. As agentic AI shifts from experimentation to production, the governance and security layer sitting between sensitive corporate data and large language models is fast becoming a fundable category in its own right.
Read more: Tech.eu
