MCP bridges agent clients to CommandLayer runtime actions.

What MCP does

• Exposes tools.
• Forwards payloads to runtime.
• Blocks private key fields.
• Returns runtime receipts.
• Supports automatic verification workflows.

What MCP does not do

• Does not hold private keys.
• Does not sign receipts.
• Is not the trust root.
• Does not make schema-valid equal verified.

Trust boundary

The proof model is identical whether an action is called through MCP or direct HTTP. MCP changes transport, not trust. Runtime still signs. Verifier still validates. MCP does not hold keys.

Transport

GET  https://mcp.commandlayer.org/health
POST https://mcp.commandlayer.org/mcp

Note: /mcp is POST-only MCP transport. GET /mcp is not expected to work.

Tools

clas.trust-verification.verify
clas.trust-verification.sign
clas.trust-verification.attest
clas.trust-verification.authorize
clas.trust-verification.approve
clas.trust-verification.reject
clas.trust-verification.permit
clas.trust-verification.grant
clas.trust-verification.authenticate
clas.trust-verification.endorse

Flow

MCP client -> mcp.commandlayer.org -> runtime.commandlayer.org -> signed receipt -> verifier -> VALID / INVALID