OpenFGA MCP: LLM-Powered Data Access & Manipulation

Key Features of OpenFGA MCP: LLM-Powered Data Access & Manipulation?

• LLM-Driven Authorization: Natural language interpretations enable dynamic permission checks and policy adjustments.
• Granular Control: Precisely scope access with fine-tuned policies, reducing over-permissions risks.
• Explainable Decisions: Automatically generate human-readable justifications for authorization outcomes.
• Ease of Integration: Supports popular MCP clients and offers flexible deployment via CLI, Docker, or source code.

Use Cases of OpenFGA MCP: LLM-Powered Data Access & Manipulation?

Unlock transformative scenarios with this technology:

• Dynamic Policy Updates: Modify authorization rules on-the-fly using conversational interfaces.
• Agent-Based Security: Enable AI systems to autonomously enforce policies while maintaining human oversight.
• Collaboration Simplified: Grant temporary, role-based access to external teams without manual intervention.
• Audit-Ready Workflows: Track policy changes and access logs for compliance reporting.

FAQ: OpenFGA MCP Insights

Q: Is it compatible with existing OpenFGA deployments?
A: Yes, MCP seamlessly integrates with your current OpenFGA infrastructure.

Q: Can I customize policy logic?
A: Absolutely—define policies in human-readable language, and let the system translate them into executable code.

Q: How secure is the LLM integration?
A: Input sanitization and role-based access controls ensure only authorized LLMs interact with sensitive data.

OpenFGA MCP

An experimental Model Context Protocol (MCP) server that enables Large Language Models (LLMs) to read, search, and manipulate OpenFGA stores. Unlocks authorization for agentic AI, and fine-grained vibe coding✨ for humans.

Built using the OpenFGA Python SDK and MCP Python SDK.

Quick Start

Requirements

• Python 3.10+
• OpenFGA

Installation

# Using pip/uv
uv pip install openfga-mcp

# From source
git clone https://github.com/evansims/openfga-mcp.git
cd openfga-mcp
make setup
source activate_venv.sh

Running

# Direct CLI usage
openfga-mcp-server --url "https://localhost:8000" --store "your-store-id"

# Using Make
make run

# With Docker
make docker-build
OPENFGA_API_URL="https://localhost:8000" OPENFGA_STORE_ID="your-store-id" make docker-run

Connecting

Connect your LLM application to the MCP server endpoint (default: http://localhost:8090).

Compatible with MCP clients including Cursor, Windsurf, Cline, Claude Desktop, and Zed.

Development

# Setup
make setup
source activate_venv.sh

# Common tasks (all run in virtual environment automatically)
make test        # Run tests
make lint        # Run linting
make type-check  # Run type checking
make check       # Run all checks

# Interactive development
make shell       # Start shell in virtual environment
make repl        # Start Python REPL
make ipython     # Start IPython REPL

# Run a custom command
make in-venv CMD="python -m openfga_mcp version"

Use Cases

1. Dynamic Access Control : LLMs interpret natural language to determine permissions based on context
2. Policy Management : Create or adjust authorization policies through conversational interfaces
3. Explainable Authorization : Provide clear justifications for access decisions
4. Policy Debugging : Diagnose permissions issues conversationally
5. Secure Collaboration : Grant temporary access with precise scope

Documentation

For detailed documentation, run:

make docs-serve

Contributing

See Contributing Guidelines for more information.

License

Apache License 2.0 - see the LICENSE file for details.