AgentMesh

The Secure Nervous System for Cloud-Native Agent Ecosystems

Identity · Trust · Reward · Governance

⭐ If this project helps you, please star it! It helps others discover AgentMesh.

🔗 Part of the Agent Ecosystem — Works seamlessly with Agent-OS for IATP trust protocol

---

Overview

AgentMesh is the first platform purpose-built for the Governed Agent Mesh — the cloud-native, multi-vendor network of AI agents that will define enterprise operations.

The protocols exist (A2A, MCP, IATP). The agents are shipping. The trust layer does not. AgentMesh fills that gap.

┌─────────────────────────────────────────────────────────────────────────────┐
│                           AGENTMESH ARCHITECTURE                            │
├─────────────────────────────────────────────────────────────────────────────┤
│  LAYER 4  │  Reward & Learning Engine                                       │
│           │  Per-agent trust scores · Multi-dimensional rewards · Adaptive  │
├───────────┼─────────────────────────────────────────────────────────────────┤
│  LAYER 3  │  Governance & Compliance Plane                                  │
│           │  Policy engine · EU AI Act / SOC2 / HIPAA · Merkle audit logs   │
├───────────┼─────────────────────────────────────────────────────────────────┤
│  LAYER 2  │  Trust & Protocol Bridge                                        │
│           │  A2A · MCP · IATP · Protocol translation · Capability scoping   │
├───────────┼─────────────────────────────────────────────────────────────────┤
│  LAYER 1  │  Identity & Zero-Trust Core                                     │
│           │  Agent CA · Ephemeral creds · SPIFFE/SVID · Human sponsors      │
└───────────┴─────────────────────────────────────────────────────────────────┘

Why AgentMesh?

The Problem

• 40:1 to 100:1 — Non-human identities now outnumber human identities in enterprises
• AI agents are the fastest-growing, least-governed identity category
• A2A gives agents a common language. MCP gives agents tools. Neither enforces trust.

The Solution

AgentMesh provides:

Capability	Description
Agent Identity	First-class identity with human sponsor accountability
Ephemeral Credentials	15-minute TTL by default, auto-rotation
Protocol Bridge	Native A2A, MCP, IATP with unified trust model
Reward Engine	Continuous behavioral scoring, not static rules
Compliance Automation	EU AI Act, SOC 2, HIPAA, GDPR mapping

Quick Start

Option 1: Secure Claude Desktop (Recommended)

# Install AgentMesh
pip install agentmesh-platform

# Set up Claude Desktop to use AgentMesh governance
agentmesh init-integration --claude

# Restart Claude Desktop - all MCP tools are now secured!

Claude will now route tool calls through AgentMesh for policy enforcement and trust scoring.

Option 2: Create a Governed Agent

# Initialize a governed agent in 30 seconds
agentmesh init --name my-agent --sponsor alice@company.com

# Register with the mesh
agentmesh register

# Start with governance enabled
agentmesh run

Option 3: Wrap Any MCP Server

# Proxy any MCP server with governance
agentmesh proxy --target npx --target -y \
  --target @modelcontextprotocol/server-filesystem \
  --target /path/to/directory

# Use strict policy (blocks writes/deletes)
agentmesh proxy --policy strict --target <your-mcp-server>

Installation

pip install agentmesh-platform

Or install with extra dependencies:

pip install agentmesh-platform[server]  # FastAPI server
pip install agentmesh-platform[dev]     # Development tools

Or from source:

git clone https://github.com/imran-siddique/agent-mesh.git
cd agent-mesh
pip install -e .

Examples & Integrations

Real-world examples to get started quickly:

Example	Use Case	Key Features
MCP Tool Server	Secure MCP server with governance	Rate limiting, output sanitization, audit logs
Multi-Agent Customer Service	Customer support automation	Delegation chains, trust handshakes, A2A
Healthcare HIPAA	HIPAA-compliant data analysis	Compliance automation, PHI protection, Merkle audit
GitHub PR Review	Code review agent	Output policies, shadow mode, trust decay

Framework integrations:

• Claude Desktop - Secure MCP tools with one command
• LangChain Integration - Secure LangChain agents with policies
• CrewAI Integration - Multi-agent crew governance

📚 Browse all examples →

The AgentMesh Proxy: "SSL for AI Agents"

Problem: AI agents like Claude Desktop have unfettered access to your filesystem, database, and APIs through MCP servers. One hallucination could be catastrophic.

Solution: AgentMesh acts as a transparent governance proxy:

# Before: Unsafe direct access
{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/me"]
    }
  }
}

# After: Protected by AgentMesh
{
  "mcpServers": {
    "filesystem": {
      "command": "agentmesh",
      "args": [
        "proxy", "--policy", "strict",
        "--target", "npx", "--target", "-y",
        "--target", "@modelcontextprotocol/server-filesystem",
        "--target", "/Users/me"
      ]
    }
  }
}

What you get:

• 🔒 Policy Enforcement - Block dangerous operations before they execute
• 📊 Trust Scoring - Behavioral monitoring (800-1000 scale)
• 📝 Audit Logs - Tamper-evident record of every action
• ✅ Verification Footers - Visual confirmation in outputs

Set it up in 10 seconds:

agentmesh init-integration --claude
# Restart Claude Desktop - done!

Learn more: Claude Desktop Integration Guide

Core Concepts

1. Agent Identity

Every agent gets a unique, cryptographically bound identity:

from agentmesh import AgentIdentity

identity = AgentIdentity.create(
    name="data-analyst-agent",
    sponsor="alice@company.com",  # Human accountability
    capabilities=["read:data", "write:reports"],
)

2. Delegation Chains

Agents can delegate to sub-agents, but scope always narrows:

# Parent agent delegates to child
child_identity = parent_identity.delegate(
    name="summarizer-subagent",
    capabilities=["read:data"],  # Subset of parent's capabilities
)

3. Trust Handshakes (IATP)

Cross-agent communication requires trust verification:

from agentmesh import TrustBridge

bridge = TrustBridge()

# Verify peer before communication
verification = await bridge.verify_peer(
    peer_id="did:mesh:other-agent",
    required_trust_score=700,
)

if verification.verified:
    await bridge.send_message(peer_id, message)

4. Reward Scoring

Every action is scored across multiple dimensions:

from agentmesh import RewardEngine

engine = RewardEngine()

# Actions are automatically scored
score = engine.get_agent_score("did:mesh:my-agent")
# {
#   "total": 847,
#   "dimensions": {
#     "policy_compliance": 95,
#     "resource_efficiency": 82,
#     "output_quality": 88,
#     "security_posture": 91,
#     "collaboration_health": 84
#   }
# }

5. Policy Engine

Declarative governance policies:

# policy.yaml
version: "1.0"
agent: "data-analyst-agent"

rules:
  - name: "no-pii-export"
    condition: "action.type == 'export' and data.contains_pii"
    action: "deny"
    
  - name: "rate-limit-api"
    condition: "action.type == 'api_call'"
    limit: "100/hour"
    
  - name: "require-approval-for-delete"
    condition: "action.type == 'delete'"
    action: "require_approval"
    approvers: ["security-team"]

Protocol Support

Protocol	Status	Description
A2A	✅ Alpha	Agent-to-agent coordination
MCP	✅ Alpha	Tool and resource binding
IATP	✅ Alpha	Trust handshakes (via agent-os)
ACP	🔄 Beta	Lightweight messaging
SPIFFE	✅ Alpha	Workload identity

Architecture

agentmesh/
├── identity/           # Layer 1: Identity & Zero-Trust
│   ├── agent_id.py     # Agent identity management
│   ├── credentials.py  # Ephemeral credential issuance
│   ├── delegation.py   # Delegation chains
│   └── spiffe.py       # SPIFFE/SVID integration
│
├── trust/              # Layer 2: Trust & Protocol Bridge
│   ├── bridge.py       # Protocol bridge
│   ├── handshake.py    # Trust handshakes
│   └── capability.py   # Capability scoping
│
├── protocols/          # Protocol implementations
│   ├── a2a/            # A2A support
│   ├── mcp/            # MCP support
│   └── iatp/           # IATP (uses agent-os)
│
├── governance/         # Layer 3: Governance & Compliance
│   ├── policy.py       # Policy engine
│   ├── compliance.py   # Compliance mapping
│   ├── audit.py        # Merkle-chained audit logs
│   └── shadow.py       # Shadow mode
│
├── reward/             # Layer 4: Reward & Learning
│   ├── engine.py       # Reward engine
│   ├── scoring.py      # Multi-dimensional scoring
│   └── learning.py     # Adaptive learning
│
├── cli/                # Command-line interface
└── sdk/                # Python SDK

Compliance

AgentMesh automates compliance mapping for:

• EU AI Act — Risk classification, transparency requirements
• SOC 2 — Security, availability, processing integrity
• HIPAA — PHI handling, audit controls
• GDPR — Data processing, consent, right to explanation

from agentmesh import ComplianceEngine

compliance = ComplianceEngine(frameworks=["soc2", "hipaa"])

# Generate compliance report
report = compliance.generate_report(
    agent_id="did:mesh:healthcare-agent",
    period="2026-01",
)

Threat Model

Threat	AgentMesh Defense
Prompt Injection	Tool output sanitized at Protocol Bridge
Credential Theft	15-min TTL, instant revocation on trust breach
Shadow Agents	Unregistered agents blocked at network layer
Delegation Escalation	Chains are cryptographically narrowing
Cascade Failure	Per-agent trust scoring isolates blast radius

Roadmap

Phase	Timeline	Deliverables
Alpha	Q1 2026	Identity Core, A2A+MCP bridge, CLI
Beta	Q2 2026	IATP handshake, Reward Engine v1, Dashboard
GA	Q3 2026	Compliance automation, Enterprise features
Scale	Q4 2026	Agent Marketplace, Partner integrations

Dependencies

AgentMesh builds on:

• agent-os — IATP protocol, Nexus trust exchange
• SPIFFE/SPIRE — Workload identity
• OpenTelemetry — Observability

Contributing

See CONTRIBUTING.md for guidelines.

License

Apache 2.0 — See LICENSE for details.

---

Agents shouldn't be islands. But they also shouldn't be ungoverned.

AgentMesh is the trust layer that makes the mesh safe enough to scale.