umpirix
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| OPENJEV_BASE_URL | Yes | Base URL of the OpenJev server that the stdio MCP server forwards to, e.g. http://localhost:8080. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
Typed decisions with probabilities. Use choice to pick one option, score to rate on a scale, noul for yes/no, and decide to ask several typed questions about the same text in one call. Branch on the returned probabilities in your own logic.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| choiceA | Classify text into exactly one of the given options. Returns the chosen option, a confidence, and a probability for every option. Use it for routing, labelling and triage instead of asking a model and parsing its reply. |
| scoreA | Rate text on an ordered scale. Returns the expected score (0 is the first level), the closest level, a confidence, and a probability for every level. Use it for urgency, severity, sentiment or quality. |
| noulA | Answer a yes/no question about text. Returns the probability that the answer is yes (0 to 1) and the yes/no answer at 0.5. Use it for gates and flags: is this spam, is this a refund request, does this need a human. |
| decideA | Ask several typed questions (choice, score, noul) about the same text in one call. Returns one typed answer per question name, each with probabilities. Use it when a decision needs more than one judgment. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 4 tools
The three primitives map cleanly to distinct output types: single-choice classification, ordinal rating, and binary yes/no. The only overlap is decide, which deliberately aggregates the other three; its description makes the batch role clear, though an agent could occasionally reach for it when a single primitive would do.
All names are short, lowercase single tokens, which is consistent in form. However, the semantics are mixed: choice and score are nouns, decide is a verb, and noul is an opaque coinage that only the description decodes.
Four tools is exactly right for a small set of classification primitives, with one tool per output type plus a convenience aggregator. Nothing feels padded or missing.
Categorical, ordinal, and binary judgments plus a multi-question batch cover the core decision space well. Minor gaps remain: no multi-label classification and no batch operation across multiple texts, but agents can work around these by composing calls.