# MITRITY

> MITRITY is an Intent-Aware Governance (IAG) platform that governs and secures autonomous AI agents. Its runtime sits beside the agents you run and validates every action they attempt against the agent's declared mission scope, before the action executes; read-only connectors bring the agents that live inside SaaS platforms into the same inventory, reviews and audit trail.

The platform has two halves. Govern: agent lifecycle approvals (assessments pre-filled from observed behavior, attested reviews with separation of duties, periodic reviews, and an approval gate that can refuse non-approved agents at runtime), EU AI Act and NIST AI RMF posture control by control, an AI risk register, AI policy statements with attestation, compliance reports with integrity hashes, evidence bundles and share links, and a tamper-evident audit trail with signed checkpoints. Secure: runtime enforcement in the customer environment (Mitrity Gateway for HTTP-based agents, MCP Sidecar for MCP-compatible agents, mesh integration, and the LLM Gateway for Anthropic, OpenAI, Azure OpenAI / AI Foundry, Google Vertex AI, Amazon Bedrock and any OpenAI-compatible host), behavioral drift detection, prompt-injection defense, DLP, credential brokering, delegation-chain governance, red teaming, model scanning and threat intelligence. SaaS connectors cover Microsoft 365 Copilot and Copilot Studio, Salesforce Agentforce, GitHub Copilot and Snowflake Cortex Agents. MITRITY AB is headquartered in Stockholm, Sweden.

## Docs

- [Quickstart](https://mitrity.com/docs/getting-started/quickstart): Deploy your first governed agent in under 10 minutes
- [Documentation](https://mitrity.com/docs): Full product documentation — gateway, sidecar, policies, security, API, integrations
- [API reference](https://mitrity.com/docs/api-reference): Interactive reference for the MITRITY REST API
- [OpenAPI specification](https://mitrity.com/openapi.yaml): Raw OpenAPI 3.1 document for the platform API
- [Integration modes](https://mitrity.com/docs/edge/modes): Gateway, Sidecar & Mesh — how agent traffic reaches MITRITY
- [SaaS-embedded agents](https://mitrity.com/docs/saas-agents/overview): Read-only connectors that discover, observe and posture-check the agents inside Microsoft 365 Copilot and Copilot Studio, Salesforce Agentforce, GitHub Copilot and Snowflake Cortex
- [Governed shell](https://mitrity.com/docs/edge/governed-shell): The gateway's sandboxed shell — judged bytes execute, credentials injected outside the command line, egress allowlisted
- [Host observer](https://mitrity.com/docs/edge/host-observer): The Linux host observer — every process execution attributed to governance from kernel process ancestry; ungoverned execution is a finding, not a heuristic

## Govern

- [Governance lifecycle](https://mitrity.com/docs/compliance/governance-lifecycle): Approve agents before they run — lifecycle statuses, assessments pre-filled from observed behavior, attested reviews with separation of duties, periodic reviews, findings that open reviews, and an approval gate that can refuse non-approved agents at runtime
- [Approval workflows](https://mitrity.com/docs/policies/approval-workflows): Hold individual actions for a human decision, with timeouts and notifications
- [Policy versions](https://mitrity.com/docs/policies/versioning): Every policy save recorded as a version; each agent shows whether it runs the version its approval named
- [Compliance frameworks](https://mitrity.com/docs/compliance/frameworks): EU AI Act and NIST AI RMF posture, control by control, computed from the governance record with every gap stated
- [AI risk register](https://mitrity.com/docs/compliance/risk-register): Inherent and residual risk per agent or program, with owners, accepted-risk justifications and a computed residual from live signals
- [AI policy statements](https://mitrity.com/docs/compliance/ai-policy): Versioned policy statements per principle, with a roster of who attested to the version in force
- [Compliance reports](https://mitrity.com/docs/compliance/reports): SOC 2, GDPR, ISO 27001, executive, inventory, AI assessment and framework reports with integrity hashes, evidence bundles, share links and schedules
- [Tamper-evident audit trail](https://mitrity.com/docs/security/audit-trail): Hash-chained audit events anchored by signed checkpoints — verify a period from the dashboard, the API or offline

## Secure

- [LLM Gateway](https://mitrity.com/docs/edge/llm-gateway): Governed model traffic to Anthropic, OpenAI, Azure OpenAI / AI Foundry, Google Vertex AI, Amazon Bedrock (OpenAI-compatible API) and any OpenAI-compatible host — DLP, model allowlists, token caps, budgets and rate limits
- [Red teaming](https://mitrity.com/docs/security/red-teaming): Governed adversarial campaigns against LLM routes and your own agent endpoints — every attempt governed, audited and budgeted, full prompts and responses kept in your environment, findings mapped to OWASP LLM Top 10, MITRE ATLAS, NIST AI RMF, the EU AI Act and ISO/IEC 42001
- [Model scanning](https://mitrity.com/docs/security/model-scanning): Scan model repositories without executing anything — pickle and PyTorch archives, safetensors and GGUF headers, hidden instructions, remote-code surfaces, unpinned revisions — and check hosted models observed at the gateway
- [AI Bill of Materials](https://mitrity.com/docs/security/aibom): The MCP servers your agents invoke and the models they load, with risk ratings and drift advisories
- [Injection detection](https://mitrity.com/docs/security/injection-detection): Prompt-injection screening before content reaches an agent
- [Credential broker](https://mitrity.com/docs/security/credential-broker): Just-in-time credential leases instead of standing secrets

## Product

- [Platform](https://mitrity.com/platform): The two halves of the MITRITY platform — Govern and Secure — pillar by pillar
- [Solutions](https://mitrity.com/solutions): Industry use cases, and the problems MITRITY solves, from approving agents to red-teaming them
- [Pricing](https://mitrity.com/pricing): Starter, Pro, and Enterprise plans

## Academy

- [MITRITY Academy](https://mitrity.com/academy): Free hands-on training in AI agent governance, with certificates

## Company

- [About](https://mitrity.com/about): Company, mission, and team
- [Contact](https://mitrity.com/contact): Sales, support, and security contacts
- [Careers](https://mitrity.com/careers): Open roles at MITRITY
- [Blog](https://mitrity.com/blog): Engineering and product notes (RSS: https://mitrity.com/feed.xml)

## Legal & Trust

- [Terms of Service](https://mitrity.com/terms): The agreement governing use of the platform
- [Privacy Policy](https://mitrity.com/privacy): How personal data is collected, used, and protected
- [Data Processing Agreement](https://mitrity.com/dpa): GDPR Article 28 processing terms and sub-processor list
- [Service Level Agreement](https://mitrity.com/sla): Uptime commitments and support levels
- [Acceptable Use Policy](https://mitrity.com/acceptable-use): What you may and may not do with the platform
- [AI transparency](https://mitrity.com/ai): How MITRITY uses AI, and how AI systems should interact with this site
- [security.txt](https://mitrity.com/.well-known/security.txt): Coordinated vulnerability disclosure — report to soc@mitrity.com

## Documentation inventory

### Getting Started

- [Quickstart](https://mitrity.com/docs/getting-started/quickstart): Get your first AI agent governed by MITRITY in under 5 minutes. Install Mitrity Gateway, connect to the control plane, and see your first event.
- [Core Concepts](https://mitrity.com/docs/getting-started/concepts): Understand the foundational concepts of the MITRITY Platform: agents, mission profiles, policies, behavioral hashes, and drift scores.
- [Demo: Mitrity Gateway](https://mitrity.com/docs/getting-started/demo-gateway): A self-contained Docker demo that runs a Claude Agent SDK agent through a scripted governance scenario with Mitrity Gateway on its MCP and built-in tools.
- [Demo: Mitrity MCP Sidecar](https://mitrity.com/docs/getting-started/demo-sidecar): A self-contained Docker demo that runs a Claude Agent SDK agent through a scripted governance scenario with Mitrity MCP Sidecar on its MCP and built-in tools.
- [Demo: Multi-Agent Governance](https://mitrity.com/docs/getting-started/demo-multi-agent): Run three agents in parallel containers, an orchestrator and two workers, and watch delegation chains, threat-intel matches and privilege boundaries emerge.

### Gateway, Sidecar & Mesh

- [Governance Coverage](https://mitrity.com/docs/edge/coverage): What MITRITY governs and what it does not: governed decision points, execution paths outside them, per-framework settings and the execution-coverage badge.
- [Deployment Guide](https://mitrity.com/docs/edge/deployment): Deploy Mitrity Gateway or Mitrity MCP Sidecar, the stdio MCP programs your agent's MCP client launches: signed binaries, container images and topologies.
- [Integration Modes](https://mitrity.com/docs/edge/modes): Two integration modes for different use cases: Mitrity Gateway for governed MCP tool aggregation and Mitrity MCP Sidecar for lightweight single-server wrapping.
- [Configuration Reference](https://mitrity.com/docs/edge/configuration): Complete YAML configuration reference for Mitrity Gateway and Mitrity MCP Sidecar. Covers all fields, defaults, and environment variable overrides.
- [Mesh Enforcement (Istio)](https://mitrity.com/docs/edge/mesh-overview): The Mitrity Mesh Authorizer brings MITRITY's decision engine to an Istio mesh, governing east-west, agent-to-agent and egress traffic per SPIFFE identity.
- [Deploy in Istio](https://mitrity.com/docs/edge/mesh-deployment): Install the Mitrity Mesh Authorizer in Istio with Helm: prerequisites, extension registration, governed namespaces, mTLS, failure modes and verification.
- [Per-Agent Identity in the Mesh](https://mitrity.com/docs/edge/mesh-identity): Bind a MITRITY agent to its Kubernetes namespace and ServiceAccount so the Mitrity Mesh Authorizer resolves each SPIFFE identity to the right agent and policy.
- [Mesh Egress Allowlist](https://mitrity.com/docs/edge/mesh-egress): Opt-in per-agent egress allowlists for the Mitrity Mesh Authorizer: govern traffic leaving the Istio mesh, with the agent's identity kept through an egress hop.
- [Mitrity LLM Gateway](https://mitrity.com/docs/edge/llm-gateway): A TLS-terminating LLM egress gateway: intent, threat-intel and DLP checks on every prompt, plus model allowlists, token caps, budgets and rate limits.
- [Governed Shell](https://mitrity.com/docs/edge/governed-shell): The gateway's sandboxed shell runs exactly the bytes MITRITY judged, with credentials injected outside the command line and output returned through DLP.
- [Host Observer](https://mitrity.com/docs/edge/host-observer): The Linux host observer: what it sees, how it attributes each process execution to governance, what leaves the host, runtime findings, installation and limits.

### SaaS-Embedded Agents

- [SaaS-Embedded Agent Governance](https://mitrity.com/docs/saas-agents/overview): Discover, observe and posture-check the AI agents that run inside GitHub, Microsoft 365, Salesforce and Snowflake, with no edge to deploy.
- [GitHub Copilot](https://mitrity.com/docs/saas-agents/github-copilot): Connect a GitHub organization through the one-click MITRITY GitHub App, or an App of your own, to discover Copilot coding agents and import their activity.
- [Microsoft 365 Copilot and Copilot Studio](https://mitrity.com/docs/saas-agents/m365-copilot): Discover a tenant's Copilot deployment and every Copilot Studio agent, import audit activity and assess Copilot Studio agents against posture checks, read-only.
- [Salesforce Agentforce](https://mitrity.com/docs/saas-agents/salesforce-agentforce): Connect a Salesforce org to discover its Agentforce agents, import setup-audit activity and read org security posture via an OAuth2 JWT-bearer connected app.
- [Snowflake Cortex Agents](https://mitrity.com/docs/saas-agents/snowflake-cortex): Connect a Snowflake account to discover its Cortex Agents, import AI-observability activity and read account security posture with key-pair JWT authentication.
- [Posture & Security](https://mitrity.com/docs/saas-agents/posture-and-security): How platform-connector posture findings work, and how MITRITY protects the read-only credentials you provide when connecting a platform.

### Policies

- [Writing Policies](https://mitrity.com/docs/policies/writing-policies): How to write MITRITY policies: how a rule is built, how it matches actions, how priority picks the rule that applies, and worked examples.
- [Enforcement Modes](https://mitrity.com/docs/policies/enforcement-modes): The three enforcement modes in MITRITY (Monitor, Alert and Enforce), and how to move safely from observing your agents to blocking actions.
- [Approval Workflows](https://mitrity.com/docs/policies/approval-workflows): Pause risky agent actions for a human decision: define hold rules, manage the approval queue, set timeouts, and get notified by email and Slack.
- [Policy Versions](https://mitrity.com/docs/policies/versioning): Every policy save is recorded as an immutable version you can inspect, compare and restore, and each agent shows whether it runs the version its approval named.

### Tools & Permissions

- [Tool Catalog](https://mitrity.com/docs/tools/tool-catalog): Browse the MITRITY tool catalog: built-in tools ship with MITRITY and custom tools are tenant-scoped; match rules, operations and categories drive governance.
- [Agent Capabilities](https://mitrity.com/docs/tools/agent-permissions): Grant per-agent capabilities in MITRITY: which tools each agent can use, with rate limits, time windows and fine-grained resource scopes beside policy rules.
- [Destination Allowlists](https://mitrity.com/docs/tools/destination-allowlists): Configure per-agent destination allowlists, the destination layer of MITRITY DLP: control which domains and endpoints agents can reach and detect exfiltration.

### Security

- [Injection Detection](https://mitrity.com/docs/security/injection-detection): Detect and respond to prompt injection against your AI agents: detection methods, probability scoring, pattern matching and false-positive management.
- [Delegation Chains](https://mitrity.com/docs/security/delegation-chains): Govern agent-to-agent delegation in MITRITY. Track delegation chains, enforce depth limits, detect privilege escalation, and audit multi-agent workflows.
- [Credential Broker](https://mitrity.com/docs/security/credential-broker): How MITRITY stores credentials, grants agents access to them, and issues time-bound leases. Includes the rotation runbook and the audit / encryption model.
- [Threat Intelligence](https://mitrity.com/docs/security/threat-intelligence): Use MITRITY's shared threat intelligence feed for cross-tenant protection: indicator types, feed sources, tenant matching and automated responses.
- [DLP Pattern Authoring](https://mitrity.com/docs/security/dlp-patterns): Define content patterns and bind them to policies with per-binding direction, action and priority: a built-in PII and credential catalog plus your own patterns.
- [AI Bill of Materials (AIBOM)](https://mitrity.com/docs/security/aibom): Your inventory of the third-party parts of an AI system, the MCP servers your agents invoke and the models they load, with risk ratings and drift advisories.
- [Model Scanning](https://mitrity.com/docs/security/model-scanning): Scan the model repositories your agents load for unsafe serialization, hidden instructions and remote code, and cross-check the models they call against policy.
- [Red Teaming](https://mitrity.com/docs/security/red-teaming): Run governed adversarial tests on your agents and LLM routes: campaigns, budgets, the oracles that decide findings and what evidence leaves your environment.
- [Tamper-Evident Audit Trail](https://mitrity.com/docs/security/audit-trail): Every audit event is hash-chained to the one before it and anchored by signed daily checkpoints. Verify a period from the dashboard, the API or offline.

### Compliance & Insights

- [Compliance Reports](https://mitrity.com/docs/compliance/reports): Generate SOC 2, GDPR, ISO 27001, executive, inventory, AI assessment and framework reports, each backed by its records, hashed, shareable and schedulable.
- [ML Insights](https://mitrity.com/docs/compliance/ml-insights): What MITRITY's ML insights measure: drift scores, per-agent risk, anomalies and the live topology, and which models run on the gateway and the control plane.
- [Compliance Frameworks](https://mitrity.com/docs/compliance/frameworks): See where your governance record stands against the EU AI Act, NIST AI RMF, ISO/IEC 42001 and 27001, GDPR, SOC 2, OWASP LLM Top 10 and MITRE ATLAS.
- [AI Risk Register](https://mitrity.com/docs/compliance/risk-register): Record each identified AI risk against an agent or your program with inherent and residual scores, an owner and a justified acceptance, seeded from assessments.
- [AI Policy Statements](https://mitrity.com/docs/compliance/ai-policy): Publish your organization's AI policy as versioned statements per principle, and keep a roster of who attested to the version in force.
- [Governance Lifecycle](https://mitrity.com/docs/compliance/governance-lifecycle): Approve agents before they run: lifecycle statuses, the organization's gate, assessments pre-filled from observed behavior, attested reviews and re-review.

### Administration

- [Roles & Access Control](https://mitrity.com/docs/administration/rbac): Team roles and access in MITRITY: what each of the four roles may do, and how agent grants and environment access narrow what a scoped user sees.
- [Environments](https://mitrity.com/docs/administration/environments): Keep Development, Staging and Production apart in MITRITY: each environment has its own agents, policies, audit logs and API keys.
- [SAML SSO](https://mitrity.com/docs/administration/saml-sso): Sign in to MITRITY through your SAML 2.0 identity provider: the settings for Okta, Entra ID and Google Workspace, setup steps and JIT provisioning.
- [API Keys](https://mitrity.com/docs/administration/api-keys): Create, scope, rotate and revoke the API keys that let your pipelines, scripts and integrations call the MITRITY API without a person signing in.
- [Billing & Plans](https://mitrity.com/docs/administration/billing): Manage your MITRITY subscription: plan tiers, scale limits, feature differences, Stripe checkout and the subscription lifecycle from upgrade to cancellation.

### API

- [API Overview](https://mitrity.com/docs/api/overview): Overview of the MITRITY REST API: authentication, base URL, versioning, rate limits, and key endpoints for managing agents and policies.

### Integrations

- [SIEM Integration](https://mitrity.com/docs/integrations/siem): Forward each runtime decision MITRITY makes to your SIEM: Splunk HEC, CEF lines over HTTPS, or a JSON webhook.
- [Framework Adapters](https://mitrity.com/docs/integrations/adapters): What a MITRITY language adapter guarantees when it wraps an agent framework: wire mapping, attestation, configuration, versioning and conformance tests.
- [Admission API](https://mitrity.com/docs/integrations/admission-api): The loopback admission API Mitrity Gateway and Mitrity MCP Sidecar serve for built-in tools: transport, authentication, endpoints, decisions and holds.

## Blog posts

- [Securing the Swarm: MITRITY Announces Native Kubernetes & Istio Mesh Governance for Agentic AI](https://mitrity.com/blog/securing-the-swarm-istio-mesh-governance) (2026-06-10): Istio secures the network but doesn't speak AI intent. The MITRITY Mesh Authorizer brings intent-based, zero-trust governance to agents in Kubernetes.
- [You Know What's In Your Software. But Do You Know What Your AI Agents Are Doing? Enter the AIBOM.](https://mitrity.com/blog/aibom-runtime-inventory-for-ai-agents) (2026-06-09): SBOMs miss the tools and MCP servers AI agents invoke at runtime. The AIBOM is a live inventory of every one, with drift advisories and risk ratings.
- [The Multi-Agent Blind Spot: How to Stop AI-to-AI Privilege Escalation](https://mitrity.com/blog/multi-agent-blind-spot-ai-to-ai-privilege-escalation) (2026-06-02): A compromised low-privilege agent can weaponize a high-privilege peer. MITRITY's delegation engine blocks that AI-to-AI privilege escalation inline.
- [Getting Started with Intent-Aware Governance: A Practical Walkthrough](https://mitrity.com/blog/getting-started-iag-walkthrough) (2026-05-21): A practical guide to AI agent governance without sacrificing development velocity, with two open-source demonstration agents you can run in minutes.
- [Why Your AI Agents Shouldn't Hold the Keys to the Kingdom](https://mitrity.com/blog/agents-keys-to-the-kingdom) (2026-05-15): Agents should borrow secrets, not own them. How just-in-time leases replace permanent credentials and shrink the attack surface to seconds.
- [Securing the Agentic Stack: Privacy-First Threat Intelligence for AI](https://mitrity.com/blog/privacy-first-threat-intel-agentic-stack) (2026-05-08): Threat intel built on IPs and file hashes doesn't fit AI agents. How MITRITY built behavior-based, privacy-first threat intelligence for autonomous agents.
- [Your Customer Service Agent Just Gave Away Your Pricing Strategy](https://mitrity.com/blog/ai-agents-as-customer-representatives) (2026-03-11): Customer-facing AI agents reach your most sensitive data, from PII to internal policies. Without governance, they leak it, one helpful response at a time.
- [When Your AI Agent Reprices Your Entire Catalog at 2 AM](https://mitrity.com/blog/ai-agents-in-ecommerce-governance) (2026-03-11): AI agents manage inventory, prices and orders on their own. Without real-time governance, one miscalculation can wipe out margins or oversell inventory.
- [Your Payment Agent Just Issued 200 Refunds to the Same Account](https://mitrity.com/blog/ai-agents-in-payment-processing) (2026-03-11): Payment agents work where the stakes are highest. Without inline governance, one compromised or misconfigured agent can drain funds or expose card data.
- [Intent-Aware Governance: Why Autonomous AI Agents Need a New Control Model](https://mitrity.com/blog/intent-aware-governance-for-the-ai-era) (2026-03-01): Security tools weren't built for AI agents that reason, adapt and act independently. Intent-aware governance is the control model designed for them.
- [Real-Time Behavioral Drift Detection for Autonomous Agents](https://mitrity.com/blog/real-time-behavioral-drift-detection) (2026-02-25): A deep-dive into the ML pipeline behind real-time behavioral drift detection, from the DriftGuard model at the edge to TrustGraph threat analysis.
- [Shared Threat Intelligence Without Shared Data: Privacy-First Cross-Tenant Protection](https://mitrity.com/blog/shared-threat-intelligence-without-shared-data) (2026-02-18): How MITRITY shares AI agent threat intelligence across tenants without exposing their data, using behavioral hashing and anonymized contributions.