Metadata-Version: 2.4
Name: council-signal-mcp
Version: 0.1.0
Summary: Signed AI measurement (13-axis, deterministic judge). Measurement, not certification; monitored containment; UNMEASURED is never wrong.
Author-email: Council of AI <hello@councilof.ai>
License: Apache-2.0
Project-URL: Homepage, https://councilof.ai
Project-URL: Repository, https://github.com/CSOAI-ORG/csoai-static-deploy2
Project-URL: Documentation, https://councilof.ai/methodology
Keywords: mcp,mcp-server,ai-measurement,compliance,signed-measurement,agent-tools,mcp-registry
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: Apache Software License
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Python: >=3.10
Description-Content-Type: text/markdown

# self-measuring-mcp — the Layer-1 discovery + forcing-function surface

<!-- mcp-name: io.github.csoai-org/council-signal-mcp -->

Title: Self-Measuring MCP Server — signed measurement an agent calls on itself.

An orchestrating agent finds us the way it finds any tool in 2026, but THIS
surface is the one that makes it autocatalytic: the agent calls the MCP server
on ITSELF, emits an Ed25519-signed measurement card, and that card feeds the
board + the badge + the training funnel. Discovery (registry, agent-card,
llms.txt, .well-known) makes us findable; the signed-card forcing function
(CI gate / marketplace badge / insurer requirement) makes us USED.

Lane note: this is a NEW surface (agent-facing plumbing that wraps the
canonical measurement), not a third copy of council_signal.py. It imports the
canonical Council measurement core (CouncilSignal/Chain) rather than reimplementing them.
