Skip to content

Plan-Lint SDK

  • Validate LLM Agent Plans

    Static analysis toolkit for checking and validating agent plans before they execute.

    Getting started

  • Policy Authoring

    Learn to write Rego policies that define security boundaries for your agents.

    Policy guide

  • MCP Integration

    Integrate plan-lint with MCP servers for enhanced security.

    MCP Integration

  • API Reference

    Comprehensive API documentation for plan-lint.

    API Reference

What is Plan-Lint?

Plan-Lint is a static analysis toolkit for validating LLM agent plans before execution. It provides a robust security layer that can prevent harmful actions, detect suspicious patterns, and enforce authorization policies - all before any code executes.

from plan_lint import validate_plan

# Your agent generates a plan
plan = agent.generate_plan(user_query)

# Validate the plan against your policies
validation_result = validate_plan(plan, policies=["policies/security.rego"])

if validation_result.valid:
    # Execute the plan only if it passed validation
    agent.execute_plan(plan)
else:
    # Handle validation failure
    print(f"Plan validation failed: {validation_result.violations}")

Getting Started

Installation

pip install plan-lint

Basic Usage

from plan_lint import validate_plan

# Validate a plan against security policies
result = validate_plan(
    plan_data,
    policies=["path/to/policies/security.rego"]
)

if result.valid:
    print("Plan is valid")
else:
    print(f"Plan validation failed with {len(result.violations)} violations:")
    for violation in result.violations:
        print(f" - {violation.message}")

Key Features

  • Static Analysis: Validate plans before execution to prevent security issues
  • Rego Policies: Use OPA's Rego language to define flexible, powerful policies
  • Integration: Works with OpenAI, Anthropic, and custom agent frameworks
  • MCP Support: Integrates with MCP servers for OAuth-aware policy enforcement
  • Custom Rules: Define your own security policies based on your specific needs

Examples

Check out our examples to see Plan-Lint in action.