Security infrastructure for autonomous AI agents.

The crisis of the future is not intelligence. It is self-authorization.

Prevent unauthorized agent actions, enforce policy before execution, and produce audit-ready evidence.

Open source MIT licensed VEX pilot-only
mcp-vanguard — live
Monitoring
Execution trace
Evidence layer

Every governed action leaves evidence behind

ProvnAI replaces blind trust with inspectable records: what the agent tried to do, who authorized it, which policy applied, and how the evidence was preserved.

Policy enforced

The action is checked outside the model before it reaches privileged tools.

Identity bound

The agent, session, and approving principal are attached to the record.

Tamper evident

Each event is linked into a witness trail that can be verified later.

evidence-capsule.json
verified
capsule_id
vex-cap-9f3a2b1e
action_hashsha256:8f41c9e2...a7d3
principalorg:provn:platform-team
policyfilesystem.read.allowlist.v3
decisionapproved / human-reviewed
witness_root0x7b8c9d0e...f2a1
status
Merkle-linked and signed
Products

Security
dimensions

Start with McpVanguard for immediate protection, or design long-term trust architecture with VEX Protocol. Two approaches to the same problem.

01
Open source
The Security Gateway

McpVanguard

An open-source security layer for AI agent tool calls. Block injection, unauthorized access, and unsafe behavior before agents reach production systems.

Explore McpVanguard
02
Pilot-only
Execution Control

VEX Protocol

Review and approve high-risk agent actions before they execute. Every decision is recorded with tamper-evident evidence your auditors can verify independently.

Explore VEX Protocol
Layer
McpVanguard
VEX
Primary role
Blocks unsafe MCP tool calls before they execute.
Controls and proves what autonomous agents are allowed to do.
Best first step
Teams deploying MCP servers or agent toolchains now.
Teams designing governed execution for regulated workflows.
Evidence output
Security logs, policy decisions, and gateway telemetry.
Cryptographically verifiable governed execution evidence.
Deployment stage
Open source and available today.
Pilot access with design partners.
Core principles

Why governed
execution matters

AI agents now execute code, call APIs, and make decisions. The infrastructure granting them this power was built for humans. The security model was not updated.

01

Execution must be inspectable.

If an AI agent can modify files, call APIs, or execute code, then its actions must be observable and reviewable. Not just in logs, but in a format that security and compliance teams can reason about.

02

Controls must sit outside the model.

Relying on the model to enforce its own constraints is a single point of failure. Policy should be enforced in a layer the model cannot rewrite, negotiate, or bypass through prompt engineering.

03

Identity should not be implicit.

When agents take action on behalf of users or systems, the chain of attribution must be explicit and verifiable. Identity should be explicit, and where needed can be rooted in hardware-backed or protocol-level trust, not inferred from session context.

04

Governance is infrastructure, not an afterthought.

Evidence, audit trails, and enforcement boundaries should be built into the architecture from the beginning. Retrofitting them onto systems designed for speed alone is expensive and usually incomplete.

Integration paths

Choose the path that matches your rollout

Start with practical security controls, threat analysis, or governed evidence depending on where your agent program is today.

01
Developer Hub

For AI Developers

Block unsafe tool calls before they execute. Ship agent features faster without security review bottlenecks.

ExploreBuild with McpVanguard
02
Architecture Hub

For Security Teams

Catch injection, exfiltration, and unauthorized access before agents reach production. Reduce incident response time from days to minutes.

ExploreView Technical Analysis
03
Governance Hub

For Governance Teams

Generate tamper-evident audit trails automatically. Pass compliance reviews without manual evidence collection or spreadsheet archaeology.

ExploreExplore VEX Evidence
Pilot access

Secure your
autonomous AI agents

We are accepting a limited number of design partners and pilot deployments for secure agent execution. Tell us what you are building.

Enterprise & Startup Pilots

Deploy controlled execution for your agent stack. We work directly with your security and platform teams.

Public-Sector & Consortium

Standard-setting conversations for secure AI in regulated environments. EU AI Act, DORA, SOC2 alignment.

Architecture Collaboration

Formal verification, ZK proof systems, and protocol-level trust. We support open academic and institutional collaboration.

We read every submission and reply within two business days.

Submissions are reviewed directly by the ProvnAI team. We do not share your information with third parties.