PIC Standard provides provenance and evidence-based gating for AI agent actions, exposed here as an MCP (Model Context Protocol) compliant server integration. This MCP setup uses the pic-standard package with the MCP feature flag to guard tool calls and provide production-grade request tracing, fail-closed behavior, and sandboxed evidence handling. By running the provided MCP demo script, you can observe how an agent proposal is evaluated against a verified contract before a tool is executed, ensuring that high-risk actions (like financial or sensitive operations) cannot proceed without demonstrable provenance and evidence. The workflow centers on the PIC verifier, trusted keyrings for evidence, and the MCP integration that enforces constraints and logs decisions for auditability.

To interact with the MCP server, install the MCP-enabled package and start the demo script. The server validates action proposals, applies fail-closed policies when evidence or provenance is insufficient, and routes validated proposals to the tool executor. You can experiment with enabling debugging via the PIC_DEBUG environment variable to see gating decisions and audit logs in real time. The MCP demo showcases how to integrate with other components (LangGraph, OpenClaw, Cordum, or HTTP bridges) while maintaining strict verification of intent, provenance, and evidence before any tool invocation.

Prerequisites

• Python 3.8+ (recommended: a virtual environment)
• Internet access to install dependencies via pip

Step-by-step installation

1. Create and activate a virtual environment (optional but recommended) python -m venv venv

   Windows

   venv\Scripts\activate

   macOS/Linux

   source venv/bin/activate

2. Install the MCP-enabled PIC Standard package with MCP support pip install "pic-standard[mcp]"

3. Run the MCP demo script to start the MCP-integrated server (or adapt to your environment) python -u examples/mcp_pic_client_demo.py

4. (Optional) Install additional optional features for extended capabilities pip install "pic-standard[langgraph]" # LangGraph integration pip install "pic-standard[crypto]" # Ed25519 signature evidence

5. Verify installation by running a quick check if the repository provides test/demo data pytest -q

Tips and notes:

• The MCP integration is designed for fail-closed operation with request tracing and DoS protection; enable PIC_DEBUG to observe gating decisions.
• Use the MCP demo as a starting point to integrate PIC Standard with your production toolchain and to validate that high-risk actions require verifiable provenance and evidence before execution.
• Environment variables such as PIC_DEBUG, PIC_LOG_LEVEL, or custom verifier configurations may be used to tailor logging and behavior in your environment.
• When deploying in production, ensure your keyring (trusted signers) is populated and kept secure, as evidence verification relies on trusted public keys with expiry and revocation support.
• The MCP flow can be extended to work with HTTP bridges or other language bindings to support non-Python integrations.