chimera-compliance 1.0.0

EU AI Act compliant AI agent compliance layer — plug into any agent framework, enforce deterministic policy guards, audit every decision.

chimera-compliance

EU AI Act compliance layer for AI agents
Plug into any agent framework. Enforce deterministic policy guards. Audit every decision.

Quickstart • Dashboard • Architecture • Integrations • EU AI Act

---

What is chimera-compliance?

chimera-compliance is a plug-in compliance layer for AI agents. It wraps any agent framework (LangChain, LangGraph, LlamaIndex, CrewAI, AutoGen) with deterministic policy guards that BLOCK, ALLOW, or ASK_HUMAN before every action executes.

Agent Action --> Policy Guard --> BLOCK / ALLOW / ASK_HUMAN --> Audit Record
                    |
              Violation report + EU AI Act compliant audit trail

Why? AI agents in enterprise are hard to audit and hard to control. chimera-compliance solves this by binding agents to a deterministic runtime guard -- every tool call, every decision goes through policy evaluation before execution.

Key properties:

• Deterministic safety -- Policy constraints enforced via rules (or Z3 formal verification with CSL-Core)
• Framework agnostic -- Works with LangChain, LangGraph, LlamaIndex, CrewAI, AutoGen, or raw LLMs
• Full audit trail -- Every decision produces a complete JSON record (Art. 12)
• Human oversight -- Confirm, override, or halt at any point (Art. 14)
• Right to explanation -- One-click HTML reports for any decision (Art. 86)

Quickstart

Install

# Core (YAML rules, no Z3)
pip install chimera-compliance

# With CSL-Core for Z3 formal verification (recommended)
pip install chimera-compliance[csl]

# Everything
pip install chimera-compliance[all]

Initialize a project

chimera-compliance init

This creates .chimera/config.yaml and a starter policy file.

Write a policy

Option A: YAML rules (no extra dependencies)

# policies/governance.yaml
rules:
  - name: manager_limit
    when: "role == 'MANAGER' and amount > 250000"
    then: BLOCK
    message: "Managers cannot approve more than $250,000"
  - name: weekend_freeze
    when: "is_weekend == 'YES' and urgency != 'CRITICAL'"
    then: BLOCK
    message: "No changes on weekends unless critical"

Option B: CSL policy (requires chimera-compliance[csl])

CONFIG {
  ENFORCEMENT_MODE: BLOCK
}

DOMAIN GovernanceGuard {
  VARIABLES {
    amount: 0..1000000
    role: {"ANALYST", "MANAGER", "DIRECTOR", "VP", "CEO"}
  }

  STATE_CONSTRAINT manager_approval_limit {
    WHEN role == "MANAGER"
    THEN amount <= 250000
  }
}

Run the agent

export CHIMERA_OPENAI_API_KEY=sk-...
chimera-compliance run

Verify a policy

chimera-compliance verify policies/governance.csl

Dashboard

chimera-compliance comes with a full-featured compliance dashboard at compliance.chimera-protocol.com.

Real-time audit monitoring -- Live feed of agent decisions with ALLOW/BLOCK/ESCALATE status, EU AI Act compliance score, and top violations.

Policy management -- Create, edit, and verify CSL/YAML policies with Z3 formal verification. View constraints, variables, and policy hashes.

Analytics -- Decision trends, block rate heatmaps, and violation frequency across all agents.

Connect Agent -- 4-step wizard to integrate compliance guard into LangChain, LangGraph, CrewAI, LlamaIndex, or AutoGen.

Agent Framework Integrations

LangChain

from langchain_openai import ChatOpenAI
from chimera_compliance.integrations.langchain import wrap_tools, ChimeraCallbackHandler

# Wrap your tools with compliance guard
guarded_tools = wrap_tools(your_tools, policy="./policies/governance.yaml")

# Or use callback handler
handler = ChimeraCallbackHandler(policy="./policies/governance.yaml")
llm = ChatOpenAI(callbacks=[handler])

LangGraph

from chimera_compliance.integrations.langgraph import compliance_node

# Add a compliance gate node to your graph
graph.add_node("compliance_check", compliance_node(policy="./policies/governance.yaml"))

CrewAI

from chimera_compliance.integrations.crewai import wrap_crew_tools

guarded_tools = wrap_crew_tools(your_tools, policy="./policies/governance.yaml")

Raw LLM (direct)

from chimera_compliance import ChimeraAgent

agent = ChimeraAgent(
    model="gpt-4o",
    api_key="sk-...",
    policy="./policies/governance.csl",
)

result = agent.decide(
    "Approve $200k marketing spend for Q3",
    context={"role": "MANAGER", "department": "MARKETING"},
)

print(result.result)       # "ALLOWED"
print(result.action)       # "Allocate $200k to digital channels"

CLI Reference

Command	Description
chimera-compliance init	Initialize project with config and starter policy
chimera-compliance run	Interactive agent with real-time reasoning display
chimera-compliance run --daemon	Pipe mode for JSON stdin/stdout
chimera-compliance stop [--force]	Graceful or immediate halt
chimera-compliance verify [POLICY]	Verify CSL policy (syntax + Z3 + IR)
chimera-compliance test [--skip-llm]	End-to-end system test
chimera-compliance audit --stats	Aggregate decision statistics
chimera-compliance audit --last 10	Recent decisions table
chimera-compliance audit --export FILE	Export records (json/compact/stats)
chimera-compliance policy new NAME	Create policy from template
chimera-compliance policy list	List and verify all policies
chimera-compliance policy simulate FILE '{"k":"v"}'	Test policy against input
chimera-compliance explain --id ID	Generate Art. 86 HTML explanation
chimera-compliance docs generate	Generate Annex IV documentation
chimera-compliance license status	Show current license tier and details
chimera-compliance license activate KEY	Activate a license
chimera-compliance license deactivate	Remove license key

EU AI Act Compliance

Article	Requirement	Implementation
Art. 9	Risk management	Formal verification (Z3) or rule-based policy evaluation
Art. 12	Record-keeping	Complete DecisionAuditRecord for every decision
Art. 13	Transparency	Right to explanation, audit queries, HTML reports
Art. 14	Human oversight	Confirm/override/halt with full audit trail
Art. 15	Accuracy & security	Deterministic policy gate, API key protection
Art. 19	Retention	Configurable retention with automatic enforcement
Art. 86	Right to explanation	Self-contained HTML reports per decision
Annex IV	Technical documentation	Auto-generated 14/19 sections

Supported Providers & Frameworks

Integration	Install	Notes
LangChain	pip install chimera-compliance[langchain]	Tool wrapper + callback handler
LangGraph	pip install chimera-compliance[langgraph]	Node guard + state checkpoint
LlamaIndex	pip install chimera-compliance[llamaindex]	Tool spec + callback handler
CrewAI	pip install chimera-compliance[crewai]	Tool wrapper for crews
AutoGen	pip install chimera-compliance[autogen]	Agent wrapper
OpenAI	pip install chimera-compliance[openai]	gpt-4o, gpt-4o-mini, o1, o3
Anthropic	pip install chimera-compliance[anthropic]	claude-sonnet-4-20250514, claude-opus-4-20250514
Google	pip install chimera-compliance[google]	gemini-2.0-flash, gemini-2.5-pro
Ollama	pip install chimera-compliance[ollama]	llama3, mistral, qwen (local)
CSL-Core	pip install chimera-compliance[csl]	Z3 formal verification

Development

git clone https://github.com/akarlaraytu/chimera-compliance
cd chimera-compliance
pip install -e ".[dev,all]"
pytest tests/ -v

Related Projects

• CSL-Core -- Chimera Specification Language compiler and runtime
• Chimera Compliance Dashboard -- EU AI Act compliance dashboard

License

Apache License 2.0 -- see LICENSE for details.

---

Building AI Governance.
_{Built by Aytug Akarlar}