Semantic trust layer that checks AI-agent actions and MCP tool calls before execution.
-
Updated
Sep 8, 2026 - Python
Semantic trust layer that checks AI-agent actions and MCP tool calls before execution.
Governed registry for the code-mode MCP platform: add an API capability by submitting one declarative JSON tool contract - no server code. CI gates each contract, publishes it to the registry, and the engine turns it into a sandboxed MCP tool.
Companion Lean 4 listings for Lean Programming for AI Agents (Award Winning Creations, 2026).
System-level research, protocols, reference implementations, conformance, and evaluation for the Ordivon agent-native computing stack.
To associate your repository with the tool-contracts topic, visit your repo's landing page and select "manage topics."