arifOS

arifOS serves as the TCP-like reliability and governance layer for MCP tool calls. It wraps MCP tool interactions in a constitutional framework to ensure ordered delivery, verifiable actions, and auditable history. This server is designed to sit between AI agents and the underlying tools, enforcing the 13-floor constitution and associated governance rules described in the project’s theory docs. With arifOS running, you can route MCP calls to any registered tool while benefiting from the integrity guarantees, audit trails, and safety constraints that the constitution provides.

Using arifOS, you interact with the MCP ecosystem through the standard MCP transport and the arifOS tooling. Tools exposed by arifOS include canonical capabilities such as anchor_session, reason_mind, search_reality, ingest_evidence, audit_rules, vector_memory, simulate_heart, critique_thought, check_vital, apex_judge, eureka_forge, seal_vault, and metabolic_loop. These tools are designed to validate intent, reason about actions, verify prerequisites, audit outcomes, and ensure that any executed operations can be sealed and rolled back if necessary. You can invoke a tool by issuing an MCP request addressed to arifOS, and the server will apply the constitution, enforce safety barriers, and return a structured, verifiable result.

For developers, arifOS provides a stable transport surface and a clear governance layer. It is designed to work with MCP-compliant clients and can surface health metrics, vault sealing status, and audit trails as part of its responses. Refer to the 13-tool reference and deployment docs for integration patterns, and leverage the provided manifest and floors to align agent behavior with constitutional expectations.

Prerequisites:

• Python 3.12+ installed on your system
• Internet access to fetch dependencies
• Access to a suitable runtime environment (local or containerized)

Option A: Install and run arifOS via Python (recommended)

1. Create a virtual environment (optional but recommended):

python3 -m venv venv
source venv/bin/activate  # Linux/macOS
venv\Scripts\activate  # Windows

2. Install the arifos package from PyPI:

pip install arifos

3. Run the MCP server (as defined in mcp_config):

python -m arifos.mcp

Option B: Install and run arifOS via Node.js (alternative)

1. Ensure Node.js is installed (v18+ recommended).
2. Install the npm package:

npm install @arifos/mcp

3. Run the MCP server (example, if the package exposes a CLI or script):

npx @arifos/mcp --standalone

Option C: Docker (if you prefer containerized deployment)

1. Pull the arifos docker image (as referenced in the manifest):

docker pull ariffazil/arifos:latest

2. Run the container (example):

docker run -d --name arifos -p 8000:8000 ariffazil/arifos:latest

3. Verify the MCP endpoint is reachable and health checks pass.

Notes:

• The exact command may vary based on your packaging and entrypoints. If you use a custom setup, point the MCP server to the correct module or script as shown in your environment.

Tips and common issues:

• Ensure Python 3.12+ is used to match the supported version for arifOS.
• If you encounter import errors, verify that the arifos package is installed in the same Python environment as the server run command.
• Set ARIFOS_LOG_LEVEL to DEBUG during troubleshooting, then revert to INFO or WARN for production.
• The MCP transport may include stdio, SSE, or streamable-http paths; configure your client to use the same transport surface.
• Review the vault and audit configurations to understand how sessions are sealed and how history is preserved.
• If using Docker, ensure container resources (CPU/m memory) are sufficient for the 13-floor constitution and audit workloads.