Agentmesh - AI agent governance middleware

Created By

angelnicolasc22 days ago

Governance middleware for AI agents: deterministic policy enforcement, cryptographic audit trails with digital signatures, DLP/PII detection, Trust Score per agent (0-100), EU AI Act compliance (Art. 9, 11, 12, 14), Agent BOM generation, and Circuit Breaker. Native support for LangGraph, CrewAI, and AutoGen.

# governance

# compliance

Overview Content Tools Comments

Overview

AgentMesh

Scan your AI agents for governance gaps. Enforce policies in production.

What is AgentMesh?

AgentMesh is a governance platform for AI agents, built in two layers:

Scan CLI (free, offline, no account) — Analyzes your codebase via AST to find governance gaps, generate an Agent BOM, and map EU AI Act requirements. Like snyk test for AI agents.
Runtime Platform (SaaS, requires account) — Middleware that intercepts tool calls in production to enforce policies, scan payloads for PII, and track agent trust. Like snyk monitor for AI agents.

Quick Start

pip install useagentmesh
agentmesh scan .
# → Governance Score: 35/100 | 8 findings | Agent BOM: 3 agents, 12 tools
# → Run `agentmesh auth login` to enable runtime governance

MCP Server (Claude Desktop)

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "agentmesh": {
      "command": "uvx",
      "args": ["agentmesh-mcp"],
      "env": {
        "AGENTMESH_API_KEY": "your-api-key"
      }
    }
  }
}

MCP Server (VS Code / Cursor)

Add to MCP settings:

{
  "mcp": {
    "servers": {
      "agentmesh": {
        "command": "uvx",
        "args": ["agentmesh-mcp"]
      }
    }
  }
}

What You Get Free (Scan CLI)

🔍 Governance Score: 0-100 score based on 33 deterministic policy rules (<2ms evaluation)
📦 Agent BOM: AST-based inventory of agents, tools, and models in your project
🛠️ Fix Snippets: Actionable remediation for every finding
📄 SARIF 2.1.0: Native GitHub Code Scanning integration
🇪🇺 EU AI Act Gaps: Detects non-compliance with Art. 9, 11, 12, 14

Supported Frameworks

Framework	Supported	Discovery
LangGraph	✅	AST-based
CrewAI	✅	AST-based
AutoGen	✅	AST-based
LangChain	✅	Standard
LlamaIndex	✅	Standard
Pydantic AI	✅	Standard

Output Formats

SARIF 2.1.0 (GitHub Code Scanning compatible)
JSON (For CI/CD integrations)
SVG Badges (For repository docs)

Benchmark Results

All measurements taken with time.perf_counter_ns(), 10,000 iterations after 1,000 warmup. Methodology & reproduction →

Policy Engine (33 deterministic rules, zero LLMs):

Scenario	P50	P99
Single rule evaluation	0.031ms	0.08ms
Full scan (33 rules)	1.84ms	3.2ms
Batch (100 tool calls)	1.79ms	2.8ms

Governance overhead is <0.2% of a typical LLM call (~800ms).

AST Framework Discovery:

Framework	Avg Latency
CrewAI	~5ms
LangGraph	~7ms
AutoGen	~9ms

Runtime Governance (SaaS Platform)

When you connect the SDK to the AgentMesh platform, you unlock runtime governance features that protect your agents in production:

🔐 DLP Runtime — Presidio-based PII/PCI scanning on tool call payloads before they hit downstream APIs
📊 Dynamic Trust Score — 0-100 EigenTrust score per agent, updated on every interaction
⚡ Circuit Breaker — Auto-suspends agents when Trust Score drops below threshold
🔐 Cryptographic Audit Trail — SHA-256 hash chain + Ed25519 digital signatures (non-repudiation)
👥 RBAC + Teams — Multi-user access control per organization
📋 EU AI Act Reports — Exportable compliance reports for regulators

The Most Advanced Agent Controls

🧭 Operational Design Domain (ODD) — Define permitted tools, rate limits, and cost caps per agent. Enforcement modes: audit, enforce, escalate
📏 Pre-Action Magnitude Limits — Pre-trade risk controls for AI agents: financial spend caps, data volume limits, blast radius constraints, and compute guardrails — validated before every action executes
🤖 Agent Identity Management — Managed credential lifecycle for non-human identities: DID provisioning, auto-rotation with grace periods, instant revocation, and ephemeral JWT support

These features require an account. Sign up free →

How We Compare

Feature	AgentMesh Scan (free)	AgentMesh Platform (SaaS)	Bifrost	Cordum
Language	Python	Python	Go	Go
Static Governance Score	✅	✅	❌	❌
Agent BOM (AST)	✅	✅	❌	❌
SARIF Output	✅	✅	❌	❌
EU AI Act Gap Detection	✅	✅	❌	❌
DLP Runtime (Presidio)	—	✅	❌	❌
Dynamic Trust Score	—	✅	❌	❌
Cryptographic Audit Trail	—	✅	❌	❌
Circuit Breaker	—	✅	❌	❌
ODD Enforcement	—	✅	❌	❌
Pre-Action Magnitude Limits	—	✅	❌	❌
Agent Identity Management	—	✅	❌	❌

Pricing

Tier	Price	Tasks/month	What you get
Free (no account)	$0	—	Scan CLI, Agent BOM, SARIF, findings
Free (with account)	$0	10,000	+ Runtime middleware, basic audit trail
Starter	$29	50,000	+ DLP runtime (Presidio)
Pro	$49	200,000	+ ODD, Magnitude Limits, Agent Identity, Trust Score, Circuit Breaker
Pro Team	$199	500,000	+ SSO, 365-day retention, 25 team members
Enterprise	Custom	Unlimited	Everything in Pro Team + BFT consensus, custom SLA, dedicated support

🔗 View Pricing Plans

Links & Resources

📜 Documentation: docs.useagentmesh.com
🌐 Landing Page: useagentmesh.com
🧩 MCP Server: Coming via Smithery/PulseMCP

AgentMesh — Governance for AI Agents

Server Config

{
  "mcpServers": {
    "agentmesh": {
      "command": "uvx",
      "args": [
        "useagentmesh"
      ],
      "env": {
        "AGENTMESH_API_KEY": "your-api-key"
      }
    }
  }
}

Recommend Servers

TraeBuild with Free GPT-4.1 & Claude 3.7. Fully MCP-Ready.

Y GuiA web-based graphical interface for AI chat interactions with support for multiple AI models and MCP (Model Context Protocol) servers.

AiimagemultistyleA Model Context Protocol (MCP) server for image generation and manipulation using fal.ai's Stable Diffusion model.

EdgeOne Pages MCPAn MCP service designed for deploying HTML content to EdgeOne Pages and obtaining an accessible public URL.

Baidu Map百度地图核心API现已全面兼容MCP协议，是国内首家兼容MCP协议的地图服务商。

Serper MCP ServerA Serper MCP Server

Tavily Mcp

BlenderBlenderMCP connects Blender to Claude AI through the Model Context Protocol (MCP), allowing Claude to directly interact with and control Blender. This integration enables prompt assisted 3D modeling, scene creation, and manipulation.

ChatWiseThe second fastest AI chatbot™

RedisA Model Context Protocol server that provides access to Redis databases. This server enables LLMs to interact with Redis key-value stores through a set of standardized tools.

CursorThe AI Code Editor