AccessIQ Agent Guard

Run AI agents in production. Safely.

Identity, authorization, and audit for every AI agent — with drop-in SDKs for the frameworks you're already using. MCP-ready, standards-conforming, and compliance-grade from day one.

The platform

Six pillars, one product

Everything ships behind a single SKU — no add-ons, no per-feature billing.

Identity for every agent

First-class registry with lifecycle states, signed agent cards (JWS), DPoP token binding, software attestation, and one-click kill switch.

Drop-in MCP gateway

Reverse proxy in front of any MCP server. We do the PDP check, redact PII / secrets, write a tamper-evident ledger row, and only then forward upstream.

ML prompt-injection blocking

Lakera Guard and Lasso Security plug straight into the gateway. Built-in heuristic scanner ships out of the box for sub-second BLOCK / FLAG verdicts.

PDP federation

Already own OPA, Cedar, or AVP? Run them alongside Agent Guard in REMOTE_ONLY, UNANIMOUS, or REMOTE_OVERRIDES mode. Universal AuthZEN v0.1 passthrough.

Continuous authorization

DPoP nonce ratchet caps stolen-token windows at ~60s regardless of JWT exp. Admin-revoke pushes via SSE — SDKs drain in-flight caches in <200ms.

Compliance built in

Hash-chained compliance ledger, ISO/IEC 42001 + EU AI Act Art. 9–15 report generators, public transparency dashboard. Show the auditor on day one.

Interoperability

Standards-conforming, not standards-leading

We ship interop, not committee work. Every flow speaks an RFC the rest of the ecosystem already supports — no proprietary protocols, no lock-in.

RFC 7591OAuth 2.0 Dynamic Client Registration

RFC 7592OAuth 2.0 DCR management

RFC 7523JWT bearer client assertion (private_key_jwt)

RFC 8693OAuth 2.0 token exchange

RFC 8707Resource indicators for OAuth 2.0

RFC 9126OAuth 2.0 PAR

RFC 9449DPoP — including §8 server-issued nonce

RFC 9728Protected Resource Metadata

Model Context ProtocolAnthropic's MCP authorization server profile

Agent-to-Agent (A2A)Google A2A protocol

OpenID AuthZEN v0.1PDP / PEP API for federation

W3C Verifiable Credentials 2.0did:accessiq:agent: credentials

Developer experience

SDKs for every agent framework

One import line gates every tool call through the PDP, hashes the arguments to the ledger, and surfaces deny reasons back to the model.

LangChain JS

@identia/agent-guard-sdk/langchain

LangChain Python

agent_guard.integrations.langchain

Vercel AI SDK

@identia/agent-guard-sdk/vercel-ai

LlamaIndex

agent_guard.integrations.llama_index

Cloudflare Agents / MCP

@identia/agent-guard-sdk/mcp-gateway

Agent Guard is included in every AccessIQ plan from Starter up.

No separate product, no separate bill — agent identity, authorization, and audit ride the same CIAM plan you already run. Capabilities unlock as you move up the tiers.