Vijil
Vijil builds trust infrastructure for enterprise AI agents: it helps organizations discover and identify agents, evaluate reliability/security/safety, enforce runtime policies, and continuously improve agents as conditions change. Its stated mission is to help enterprises use AI agents that are verifiably reliable, secure, and safe.
At a Glance
- Enterprises deploying business-critical AI agents
- Regulated industries and organizations with compliance obligations
- Developers and ML/AI engineering teams
- Application security, CISO, governance, risk, and compliance teams
- +2 more
AI Tools by Vijil
(1)
Vijil
Enterprise AI Agent Trust Platform
Discussions
No discussions yet
Be the first to start a discussion about Vijil
Latest News
Vijil launches Diamond Adaptive Red Team for agents to test multi-agent systems
Leading AI researcher Tim G. J. Rudner joins Vijil as Chief Scientist
Vijil launches platform enabling AI agents to adapt to attacks and failures, including Darwin
Vijil raises $17 million, bringing total funding to $23 million, and is named a Gartner Cool Vendor
Products & Services
Agent discovery and identity inventory. Scans cloud, Kubernetes, VMware, endpoint, browser, and GitHub surfaces; identifies frameworks, models, tools, endpoints, and authentication, then lets organizations register agents and assign SPIFFE-based identities.
Trust evaluation and verification for agents. It compiles an agent's purpose, personas, policies, tools, and system prompt into bespoke probes, runs reliability/security/safety tests and adaptive red-team campaigns, and produces Trust Scores, transcripts, and compliance-ready reports.
Runtime control and open-source guardrails engine (Apache 2.0). It checks model inputs, model calls, tool calls, and outputs against policy and can allow, steer, or deny actions; it also provides monitoring and decision traces.
Agent evolution/improvement module that learns from production telemetry and Diamond/Dome findings, searches variants within constraints, and proposes evidence-backed changes to prompts, configuration, or source code for review and re-evaluation.
Market Position
Vijil positions itself as vendor-independent trust infrastructure spanning the full agent lifecycle—discover, identify, evaluate, defend, monitor, and evolve—rather than only an LLM benchmark, a static guardrail, or a model-provider safety feature. Its differentiation is bespoke, evidence-backed testing of the whole agent (including tools and policies), runtime enforcement, and closed-loop improvement from production telemetry. Comparable categories include AI security/guardrails, LLM evaluation, agent observability, and AI governance vendors.
Leadership
Founders
Vin Sharma
Founder and CEO. Previously ran engineering at Foursquare, led the deep-learning platform team for Amazon SageMaker at AWS, built Intel's data-center platform for BMW self-driving cars, and managed HP's Linux software business; Vijil credits him with 30 years across cybersecurity, open source, and AI/ML, 20 products, 10 patents, and five papers.
Subho Majumdar
Co-founder and former Head of AI at Vijil. He previously was a senior scientist in security research at Splunk and in Data Science and AI Research at AT&T Labs; he holds a PhD in Statistics from the University of Minnesota and has worked in trustworthy AI and statistical machine learning.
Zdravko Pantic
Co-founder. Previously a Senior Software Development Manager at Amazon Web Services.
Executive Team
Vin Sharma
Founder & CEO
Former GM and Director of Engineering at AWS SageMaker; previously engineering leadership at Foursquare, Intel, and HP.
Radina Mihaleva
Head of Business Development
Former COO at Astronomer; helped scale Lacework from $1 million to $100 million ARR; 20 years in cybersecurity go-to-market.
Board of Directors
Founding Story
Vijil was started to address the gap between the promise of autonomous/generative AI agents and the evidence enterprises need before putting them into business-critical or regulated workflows. The founders' initial vision was a trusted operating system/infrastructure for developing and deploying agents, with measurable reliability, security, and safety rather than informal benchmark or 'vibe' checks.
Business Model
Revenue Model
Freemium hosted SaaS priced per agent, plus annual subscriptions for deployments in a customer's VPC, datacenter, on-premises, or air-gapped network. Vijil says it does not meter per evaluation or per token; professional services/guided pilots are also offered.
Pricing Tiers
Dome control engine and Discover endpoint run on the customer's own hardware; Dome is Apache 2.0.
One agent on console.vijil.ai; includes a baseline evaluation per week, one Quick adaptive run per month, one enforced agent, and one Darwin pull request.
Monthly or annual; supports selected agents, evaluation on every change, adaptive runs, enforcement, and Darwin proposals. No per-evaluation or per-token meter.
Runs in a VPC, datacenter, or air-gapped network; unmetered by agent count or tokens.
Target Markets
- Enterprises deploying business-critical AI agents
- Regulated industries and organizations with compliance obligations
- Developers and ML/AI engineering teams
- Application security, CISO, governance, risk, and compliance teams
- Businesses using agentic transactions or third-party agents
- Cloud/platform teams operating agents across heterogeneous infrastructure
- Testing and certifying enterprise AI agents before production
- Continuous evaluation in CI/CD and after agent/model/tool changes
- Runtime guardrails and policy enforcement for agents in production
- Discovering shadow/rogue agents and building an enterprise agent registry
- Hardening agents against prompt injection, jailbreaks, data leakage, tool misuse, and adversarial attacks
- Governance, audit, and compliance evidence for regulated or business-critical AI
- SmartRecruiters
- Autonoma
- Brainforest
- DuploCloud