You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
chore(spec): propose unified MCP tool_registry .tri upstream in t27 #7
chore(spec): propose unified MCP tool_registry .tri upstream in t27
Info-drop, async-only.
Context
We drafted a .tri specification for a unified MCP tool registry: a small SSOT for Tool { name, version, args_schema, returns_schema, side_effect_class } that any MCP server can consume.
The spec lives at gHashTag/t27 under specs/mcp/tool_registry.tri (ring-105-004). A rust STUB target is planned as gen/rust/tool_registry.rs (auto, no hand-edit).
Intent: schema-first, so that a downstream MCP server (including trios-mcp) can align its tool descriptors with the SSOT before any code lands.
Three optional questions
Are the side_effect_class values (pure / read / write / net / exec) aligned with the tri-flavoured MCP convention used in trios-mcp, or should they be renamed?
Should args_schema reference a JSON Schema string, or a first-class .tri type node?
Is there an existing trios-mcp issue tracking a shared tool schema, to avoid duplication?
Activity
feat(specs/mcp): add unified MCP tool_registry .tri (rust STUB target)