jev-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| JEV_KEY | No | Alias for TYPESAFE_API_KEY. | |
| JEV_MODEL | No | Pin a version (e.g. jev-1.13.0) in production. | jev-latest |
| JEV_API_KEY | No | Alias for TYPESAFE_API_KEY. | |
| JEV_BASE_URL | No | Override to route through a gateway. | https://api.typesafe.ai/v1/systemone |
| TYPESAFE_API_KEY | No | Your Jev key. Required; may also be provided via JEV_API_KEY or JEV_KEY. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| jev_classifyA | Pick exactly one of up to 255 labelled options for the given state. Returns the winning option, per-option probabilities, and a confidence. Use for routing, categorization, intent detection. |
| jev_scoreA | Place the state on an ordered 2–10 level scale you define. Returns a (possibly fractional) score and the full distribution. Use for risk, urgency, quality, severity, fit. |
| jev_checkA | Answer a yes/no question about the state as a calibrated probability from 0 to 1. Use for gates, filters, and guardrails (e.g. 'Is this safe to auto-approve?'). |
| jev_gateA | Screen an action (e.g. a shell command or file edit) before an agent runs it. Returns a risk score, whether it touches production/data, and a recommendation: allow / confirm / block. |
| jev_decideA | The full Jev call: send a state and a map of typed questions, get all answers in one round trip. Each question is { type: 'choice'|'score'|'noul', instructions, criteria }. choice.criteria is a {key:desc} map; score.criteria is an ordered string array; noul has no criteria. |
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 5 tools
Each tool targets a distinct decision primitive: categorical selection, ordinal scoring, binary yes/no, action screening, and a combined multi-question call. The only mild ambiguity is that jev_decidee is a superset of the individual question types, and jev_checck and jev_gate both deal with safety-related decisions, but their outputs are clearly different.
All tools follow the same jev_<verb> convention with lowercase snake_case names and no mixed styles. The verb choice is consistent with each tool's single responsibility, making the set predictable.
Five tools is a well-scoped size for this server: four primitives plus one combined/everytive endpoint. Each tool earns its place and there is no padding or bloat.
The suite covers the visible decision types (choice, score, yes/no, gate) and the decide tool adds a batched path that also supports the noul type. There are no obvious dead ends or missing operations for the stated purpose.