Map / AI Identity / Reva
Reva
reva.ai ↗ · LinkedIn ↗ · Open on the map
AI IdentityMCP & LLM GatewaysRuntime & GuardrailsObservability & GovernanceAI Data Governance
Capabilities
Continuous adaptive authorization platform for humans and AI agents, positioned as an Authorization Management Platform (AMP) rather than an IdP. Four modules: Policy Control Tower (no-code and code policy authoring, testing and simulation, versioning, approvals, policy certification), Reva Trust Gateway (runtime enforcement point for apps, APIs, Kubernetes, microservices, cloud IAM, and agents), Policy Intelligence (explainable decisions and behavior analytics), and Data Fabric (access graph and entitlement visibility). Its Intent & Behavior Based Access Control (IBAC) model evaluates what an agent is trying to accomplish — not just who it is — before each action, and extends to agent registries, MCP server and tool invocation authorization, and permission-aware RAG retrieval. Ships an AWS-specific authorization control plane spanning IAM, SCPs, Verified Permissions, and Bedrock AgentCore policies.
Sensors / integration
Distributed Trust Gateway decision points called inline by applications, APIs, Kubernetes services, agent frameworks, and MCP servers; real-time risk and session signals ingested over the Shared Signals Framework (SSF) and CAEP; continuous behavioral telemetry on agent actions, tool usage, delegation chains, and execution flows; automated discovery and registration of GenAI workloads, RAG pipelines, agents, and MCP servers to surface shadow AI; agent and MCP server registries plus entitlement data pulled from cloud IAM, SaaS, data platforms, and vector stores into an access graph; integrates with open policy engines (Cedar, OPA, AVP, AuthZen) and SPIFFE/SPIRE identities.
Protections
Per-action allow/deny decisions at an average sub-40ms latency instead of static roles or long-lived scopes; intent validation that checks a requested action against approved objectives before it executes; goal-drift and excessive-agency detection via a trained LLM-as-Judge and self-learning models (claimed up to 98% drift-detection accuracy); Access Clipping that removes standing privileges (cited at ~80% reduction); human-in-the-loop approval workflows for high-risk actions; fine-grained policy on which agents or apps may invoke which MCP servers and tools; permission-aware retrieval that filters vector-store results (OpenSearch today, Pinecone and Milvus planned) to the requester's entitlements before they reach the model; immutable decision logs, policy version history, and a policy library mapped to OWASP LLM Top 10, OWASP Agentic, NIST AI RMF, and the EU AI Act.
Differentiator
Founded in 2025 out of Los Angeles by the Saviynt founding bench — CEO Amit Saha (Saviynt co-founder and ex-Chief Growth Officer), CSO Yash Prakash (ex-Saviynt CSO/COO), and CTO Tushar Agarwal (ex-Saviynt engineering leader) — rebuilding authorization, not identity governance, for the agentic era. The bet is IBAC: authorization past identity and role, continuously re-evaluated against intent, behavior, context, and risk, with authorization treated as one control plane across AI agents, apps, and data rather than agent controls bolted onto legacy IAM. Built on open policy-as-code standards (Cedar, AWS Verified Permissions, OPA, AuthZen) to avoid engine lock-in.
Team & funding
- Founders
- Amit Saha, Yash Prakash, Tushar Agarwal
Other AI Identity companies
8Layers · AccuKnox · Act Security · Aembit · agen.co · Agentic Fabriq · aizome · AKA Identity · Akeyless Security · Alter · Andromeda Security · Apono · See all →
Last updated · Get the “Listed on” badge