Vijil Trust Infrastructure for AI Agents

Vijil is a California company founded in 2023 by former AWS senior leaders. It builds trust infrastructure that lets enterprises hand tasks to teams of AI agents, which earn autonomy by proving their identity, resilience and fidelity, and has raised 23 million USD from BrightMind Partners, Mayfield and Gradient.

The platform runs a trust loop of four products. Discover inventories agents across code, cloud and endpoints and gives each a SPIFFE-based identity. Diamond verifies agents with bespoke test suites and adaptive adversarial red teaming scored on tests the agent never trained on. Dome, which is open source, guards LLM and tool calls and limits data access and disclosure with two lines of code. Darwin, in development with design partners, traces failures and proposes hardened agent versions as pull requests. Developers work through a Python SDK and CLI, a Claude Code plugin or a hosted console, and results support evidence for the EU AI Act, NIST AI RMF and ISO 42001.

• Discover: inventories agents across code, cloud and endpoints with SPIFFE-based identities.
• Diamond: bespoke test suites and adaptive adversarial red teaming on unseen tests.
• Dome: open-source guard for LLM and tool calls that limits data access and disclosure.
• Darwin: traces failures and proposes improved agents as pull requests for approval.
• Developer tooling: Python SDK and CLI, Claude Code plugin and hosted console.
• Compliance evidence: reporting aligned with the EU AI Act, NIST AI RMF and ISO 42001.

Vietnamese software teams building agents for e-commerce or customer service can add Dome guards and Diamond test suites before release. HR platforms can verify recruitment agents for data disclosure risks, as SmartRecruiters does. Banks piloting agentic commerce can keep an inventory and identity for every agent in production.

Teams can start free, with no sales call or credit card, and use the open-source Dome guard without an account. Testing on unseen adversarial suites gives a more honest picture of agent resilience than benchmarks the agent was tuned on. In Vijil's published benchmark, Dome improved 10 of 11 models on their worst 10% of responses, with a median gain of 17.6 points. SmartRecruiters reports using Vijil to cut time-to-trust for agents.

Enterprises are moving from single chatbots to fleets of agents that act on business systems, and analysts now track agentic AI trust, risk and security management as its own category. Agent identity, continuous testing and runtime guards are becoming prerequisites for granting agents more autonomy, including in Southeast Asian banking and commerce.

Founded in 2023 by former AWS senior leaders and based in California, Vijil raised a 17 million USD Series A in November 2025 led by BrightMind Partners with Mayfield and Gradient, for 23 million USD in total. It was named a 2025 Gartner Cool Vendor, listed in the 2025 CB Insights AI 100, won a 2026 Intellyx Digital Innovator Award and states SOC 2 Type II attestation.