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
I have verified this would not be more appropriate as a feature request in a specific repository
I have searched existing discussions to avoid duplicates
Your Idea
Problem
Current MCP ecosystem lacks a standardized, token-efficient syntax for users
to explicitly request tool execution within prompts. Developers rely on
natural language ("please search for...") or framework-specific conventions,
leading to:
Inconsistent behavior across models (Claude, GPT, Llama interpret hints differently)
Higher token costs for explicit tool requests
No deterministic way to force execution vs. let model decide
Proposal
Compact inline trigger syntax for user-to-model tool hints:
||MCP@search:{"q":"weather in Tokyo"}||
||MCP@compute:{"a":157,"b":189}||
||MCP@sum:[1,2,3,4,5]||
EXEC:||MCP@log:{"event":"user_login","ts":1704067200}||
System Prompt Example
MCP-Services Tool Trigger:
If prompt has: ||MCP@service:{json}|| - call if necessary;
OR: EXEC:||MCP@service:{json}|| - force call;
call specified service by std API tool call with json params.
Syntax Specification
Delimiter: || (double pipe, rare in natural text (single token))
Namespace: MCP@ (prevents collision with wiki-links [[...]])
Service slot: service (maps to MCP tool name)
Payload: valid JSON (object or array)
Benefits
Token efficiency: ~5 tokens vs 15-30 for natural language equivalent
Unambiguous: || not used in Markdown/HTML/LaTeX; MCP@ namespace isolates from wiki syntax
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
-
Pre-submission Checklist
Your Idea
Problem
Current MCP ecosystem lacks a standardized, token-efficient syntax for users
to explicitly request tool execution within prompts. Developers rely on
natural language ("please search for...") or framework-specific conventions,
leading to:
Proposal
Compact inline trigger syntax for user-to-model tool hints:
Auto-detect (model decides necessity):
||MCP@service:{json}||
Mandatory force call:
EXEC:||MCP@service:{json}||
Examples
||MCP@search:{"q":"weather in Tokyo"}||
||MCP@compute:{"a":157,"b":189}||
||MCP@sum:[1,2,3,4,5]||
EXEC:||MCP@log:{"event":"user_login","ts":1704067200}||
System Prompt Example
MCP-Services Tool Trigger:
If prompt has: ||MCP@service:{json}|| - call if necessary;
OR: EXEC:||MCP@service:{json}|| - force call;
call specified service by std API tool call with json params.
Syntax Specification
||(double pipe, rare in natural text (single token))MCP@(prevents collision with wiki-links[[...]])service(maps to MCP tool name)Benefits
||not used in Markdown/HTML/LaTeX;MCP@namespace isolates from wiki syntaxEXEC:prefix removes model's discretionReference Validation
Tested with: Kimi (Moonshot), GPT-4, Claude 3.5 Sonnet, Llama 3.1
Success rate: >95% for trigger recognition when specified in system prompt.
Scope
Beta Was this translation helpful? Give feedback.
All reactions