Markovian Protocol / Model Context Protocol

Model Context Protocol, with verifiable provenance Live

The Model Context Protocol is a standard socket that connects AI agents to outside tools: a server publishes a menu of tools in plain language, and an agent calls what fits. The Markovian Protocol runs one. It stamps data into a public, witnessed transparency log and returns a provenance stamp any MCP client can verify.

How COMMIT plugs in

A stamp tool on a live MCP server

Public Streamable HTTP at https://api.markovianprotocol.com/mcp/.

markovian_stamp

Stamps a piece of data into the public transparency log.

markovian_verify

Checks a merkle root against the public verifier and returns whether the log holds a matching record.

markovian_trace

Walks a stamp's derived_from lineage by merkle root and verifies every node and edge. Returns a map of the graph, not a verdict.

Try it

Add the server to an MCP client

No install. Add the endpoint to a client configuration:

mcpServers, hosted HTTP

{ "mcpServers": { "markovian": { "type": "http", "url": "https://api.markovianprotocol.com/mcp/" } } }

The same tools run locally over stdio, from the reference server in the protocol repository.

mcpServers, local stdio

{ "mcpServers": { "markovian": { "command": "python3", "args": ["markovian_mcp.py"] } } }

Why the anchor matters

A committed result is tamper-evident after the fact

A commitment fixes a result to a time in a public log whose checkpoints are cosigned by independent witnesses and anchored to Bitcoin; altering or backdating it undetected would require breaking SHA-256 or rewriting Bitcoin. The checking party never has to trust, or contact, the producer.