agentward-ai/agentward

Overview

Agentward is an open-source permission control plane for AI agents. It intercepts every tool call made by an agent, applies permission policies, and logs all calls for audit. Written in Python, it helps developers enforce security rules on agent actions.

Best for

Best for
Teams that need to enforce and audit permissions in agentic workflows

Use cases

• Enforce access control for agent tool calls
• Audit tool usage in agent workflows
• Scan agent behavior for policy violations

How to use

Install

pip install agentward

Tools exposed

• payload_fragmentation
• privilege_escalation
• circular_chaining
• exfiltration_sequence
• tool_name
• chaining_source
• chaining_target
• requires_policy_feature
• protected_paths
• path_traversal
• scope_creep
• skill_chaining
• pii_injection
• boundary_violation
• prompt_injection

Tested with

Claude Desktop, Claude Code, Cursor, Windsurf, VS Code, ChatGPT

Notes

Agentward is an open-source permission control plane for AI agents. It intercepts every tool call made by an agent, applies permission policies, and logs all calls for audit. Written in Python, it helps developers enforce security rules on agent actions.

18 stars on GitHub. Last updated 2026-04-29.

Use cases

• Enforce access control for agent tool calls
• Audit tool usage in agent workflows
• Scan agent behavior for policy violations

Pros

• Open source and self-hostable
• Provides a clear audit trail of every tool call
• Lightweight Python implementation

Cons

• Low GitHub stars (18) suggests early stage or low adoption
• Limited community support and documentation
• Tied to Python ecosystem, not language-agnostic

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.