London's Archestra.AI closes $10M seed to build the open-source security layer between LLMs and enterprise data

Connecting AI agents to sensitive enterprise data poses a specific security risk. When an agent can access internal systems, process content from outside the organisation and send information outward, a prompt injection in an incoming document or message can trigger data exfiltration without any user action. This vulnerability has been demonstrated on widely used platforms, and enterprises that have experienced such incidents are reluctant to rely solely on the safety controls built into LLM APIs.

Archestra.AI, founded in London in 2025 by Matvey Kukuy, Joey Orlando, and Ildar Iskhakov, has developed an open-source MCP platform that establishes a deterministic control layer between LLMs and enterprise data. The rules it enforces are fixed and predictable; they are independent of the model's judgement or the wording of a prompt. The platform is compatible with Claude, GPT-4, Gemini and open-source models. It includes a private MCP registry that governs which tools agents can access, a built-in RAG stack that runs entirely within the customer's own infrastructure, and a Kubernetes-native orchestrator for multi-team deployments. The platform is self-hostable and fully open-source.

In June 2026, Archestra closed a $10M seed round led by 20VC, with Visible Ventures and Tenacity Capital participating alongside angel investors. The round brings total funding to $13.5M, which will be used for go-to-market expansion and accelerating enterprise deployments.

Sources: Archestra.AI | Crunchbase

Founders: Joey Orlando, Matvey Kukuy, Ildar Iskhakov