Overview & result
This proof of concept discovers a simple MCP server's tool signatures, then uses Jev Choice and Noul questions to select a tool and extract primitive-typed arguments from a command. The author explicitly does not claim support for arbitrary MCP schemas.
From the original post
Zero general-purpose LLM calls — anywhere in this pipeline, not for picking the tool, not for reading a parameter's value out of a sentence, not for anything. This is a proof of concept for a different way to do MCP tool dispatch: given a simple MCP server — one whose tools take primitive-typed parameters — this script dynamically discovers its tool signatures at runtime and uses Jev to route a plain natural-language command to the correct tool and extract its arguments straight out of the sentence, using only Jev's typed primitives (Choice / Noul) instead of free-text generation.
How Jev fits in the loop
- Discover available MCP tool signatures at runtime.
- Use Jev Choice to select a tool from the bounded registry.
- Use typed Jev questions to extract primitive arguments from the command.
- Validate and dispatch the selected MCP call in code.
How to reproduce
- Inspect the source implementation at https://github.com/abhishekashokvkumar/jev-mcp-dispatcher.
- Configure local Jev API credentials and register required tool/model bindings.
- Execute test prompts and observe decision latency and routing outputs.
Why this build matters
Demonstrates typed tool selection and parameter extraction without a general-purpose LLM call, within the author's stated simple-schema scope.
Limitations
- The README says this is not "works with any MCP server": it needs a simple server whose tools take primitive-typed parameters.
- A command is split into at most 5 independent actions.
- It is a proof of concept and expects input text in a fairly specific shape.