What is the key difference between Agentgraph and Agentward?

Agentgraph belongs to the Security category with local stdio execution, while Agentward belongs to the Security category with local stdio execution.

How do I install Agentgraph or Agentward in Claude Desktop or Cursor?

Copy the mcpServers JSON configuration snippet provided on this page into your claude_desktop_config.json or ~/.cursor/mcp.json file and restart your AI client.

Are Agentgraph and Agentward free or paid?

Agentgraph operates under a Free / Open Source model (No auth required). Agentward operates under a Free / Open Source model (No auth required).

Which MCP server has more tools and community activity?

Agentgraph lists 10 tools with 2 views and 4 GitHub stars. Agentward lists 14 tools with 4 views and 19 GitHub stars.

Can I use both Agentgraph and Agentward together in the same MCP host?

Yes! MCP hosts like Claude Desktop, Cursor, Windsurf, Cline, and VS Code support configuring multiple servers under the mcpServers object simultaneously.

Side-by-Side Model Context Protocol Comparison

Agentgraph vs Agentward

In-depth architectural comparison of the Agentgraph and Agentward MCP servers. Compare execution transports, security boundaries, tool capabilities, quality scores, and ready-to-paste client installation snippets for Claude, Cursor, Windsurf, and VS Code.

At a Glance & Executive Verdict

Agentgraph

Security · Local stdio

Quality: 63/100 (Good) | Auth: No auth required

Agentward

Security · Local stdio

Quality: 63/100 (Good) | Auth: No auth required

Verdict Summary: Choose Agentgraph if you need specialized Security tools running via a local process. Choose Agentward if your workspace requires Security integration with local subprocess execution. Both servers can be configured concurrently in your client's mcpServers manifest.

Which MCP Server Should You Choose?

Choose Agentgraph when:

You need dedicated capabilities in the Security domain.
You prefer local stdio subprocess transport architecture.
Your security boundary fits: No auth required (Free / Open Source).
You have access to required keys: DATABASE_URL, POSTGRES_PASSWORD, REDIS_URL, JWT_SECRET.
Primary tools included: verify_trust, lookup_identity, check_interaction_safety.

Explore Agentgraph Details

Choose Agentward when:

You need dedicated capabilities in the Security domain.
You prefer local stdio subprocess transport architecture.
Your security boundary fits: No auth required (Free / Open Source).
Primary tools included: read_file, read_text_file, read_media_file.

Explore Agentward Details

Feature & Specification Comparison

Specification	Agentgraph agentgraph-co Security	Agentward agentward-ai Security
Summary	Trust verification and security scanning for AI agents. Checks security posture of third-party MCP servers and tools with signed attestations (Ed25519/JWS) before interaction.	Permission control plane for AI agents. MCP proxy that enforces least-privilege YAML policies on every tool call, classifies sensitive data (PII/PHI), detects dangerous skill chains, and generates compliance audit trails. Supports stdio and HTTP proxy modes.
Category & Scope

Tools & Capabilities Breakdown

Agentgraph Tools (10)

verify_trust

Verify an entity's trust score on AgentGraph. Returns JSON with trust_score (0.0-1.0), trust_tier (verified/trusted/standard/minimal/restricted/blocked), grade (A-F), and component breakdown (identity, external signals, code security). Read-only, no auth required. Use before interacting with unknown agents to assess risk.

lookup_identity

Look up an entity on AgentGraph by DID or display name. Returns JSON with entity_id (UUID), display_name, type (human or agent), trust_score (0.0-1.0), trust_tier, capabilities array, DID (did:web:...), and bio. Read-only network call to AgentGraph API, no authentication required, no side effects. Typical response time under 500ms. Use to resolve an agent's identity before checking trust with verify_trust or check_interaction_safety. Returns null fields if entity not found.

check_interaction_safety

Check if it is safe to interact with another agent based on trust scores. Returns JSON with: safe (boolean), risk_level (low/medium/high), trust_score (0.0-1.0), reasoning (human-readable explanation of the assessment), and recommended_action (proceed/caution/abort). Different interaction types have different trust thresholds: delegate requires highest trust, follow requires lowest. Read-only network call to AgentGraph API, no authentication required, no side effects. Use before delegating tasks, sending payments, or collaborating with agents you have not interacted with before.

Ready-to-Paste Client Configurations

Paste either (or both) of these JSON server blocks into your client config file (e.g. claude_desktop_config.json or ~/.cursor/mcp.json).

Agentgraph Configuration

mcpServers (Claude Desktop / Cursor)

{
  "mcpServers": {
    "agentgraph-co-agentgraph": {
      "command": "uvx",
      "args": [
        "agentgraph-trust"
      ],
      "env": {
        "DATABASE_URL": "YOUR_DATABASE_URL_HERE",
        "POSTGRES_PASSWORD": "YOUR_POSTGRES_PASSWORD_HERE",
        "REDIS_URL": "YOUR_REDIS_URL_HERE",
        "JWT_SECRET": "YOUR_JWT_SECRET_HERE"
      }
    }
  }
}

Agentward Configuration

mcpServers (Claude Desktop / Cursor)

{
  "mcpServers": {
    "agentward-ai-agentward": {
      "command": "uvx",
      "args": [
        "agentward"
      ]
    }
  }
}

Frequently Asked Questions

Agentgraph is categorized under Security and uses a local stdio subprocess. In contrast, Agentward belongs to Security using local stdio subprocess. Select Agentgraph when you need capabilities focused on security and Agentward when you require tools for security.

More alternatives to Agentgraph More alternatives to Agentward Security category hub