agentshield

Multi-layer agent defense framework for AI security

Part of the AumOS open-source agent infrastructure portfolio.

Features

• SecurityPipeline scans agent input, output, and tool calls through an ordered chain of scanners with configurable severity thresholds and BLOCK/WARN/LOG actions
• Eight built-in scanners: regex injection detection, PII detection, credential detection, output safety, tool call validation, behavioral checking, output validation, and tool call integrity checking
• Scanner ABC makes it straightforward to write and register custom scanners; load them from a shield.yaml config file or inject them at runtime
• Phase-aware dispatch — each scanner declares which phases (INPUT, OUTPUT, TOOL_CALL) it runs in, so no scanner is invoked unnecessarily
• Cumulative findings tracked across the session with JSON, Markdown, and HTML report generation
• Adapters for LangChain, CrewAI, AutoGen, OpenAI Agents, and MCP that wrap existing agent code with a single decorator or context manager
• OWASP ASI Top 10 category mapper tags each finding with its corresponding ASI category for structured vulnerability reporting

Quick Start

Install from PyPI:

pip install agentshield

Verify the installation:

agentshield version

Basic usage:

import agentshield

# See examples/01_quickstart.py for a working example

Documentation

• Architecture
• Contributing
• Changelog
• Examples

Enterprise Upgrade

For production deployments requiring SLA-backed support and advanced integrations, contact the maintainers or see the commercial extensions documentation.

Contributing

Contributions are welcome. Please read CONTRIBUTING.md before opening a pull request.

License

Apache 2.0 — see LICENSE for full terms.

Part of AumOS — open-source agent infrastructure.