JevMCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| JEV_MODEL | No | Model for the primary provider (TypeSafe / OpenRouter default). | jev-latest / typesafe/jev-1.13 |
| JEV_API_URL | No | Endpoint URL for the primary provider. | |
| JEV_FALLBACK | No | If 'true', if the chosen provider fails, retry on the other one (needs both keys). | off |
| JEV_PROVIDER | No | Which key to use: 'typesafe' or 'openrouter'. Auto picks TypeSafe if its key is set, otherwise OpenRouter. | auto |
| JEV_MAX_ITEMS | No | Refuses a call that would send more than this many items (safety cap). | 500 |
| JEV_CHUNK_CHARS | No | Files larger than this are split into path#N chunks. | 60000 |
| JEV_CONCURRENCY | No | Max requests to Jev in flight at once. | 16 |
| TYPESAFE_API_KEY | No | TypeSafe API key. Set this or OPENROUTER_API_KEY. | |
| OPENROUTER_API_KEY | No | OpenRouter API key. Set this or TYPESAFE_API_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": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| jev_askA | Answer typed questions about files via the Jev classifier without reading them into your context. noul=yes/no (criteria optional); choice: criteria {option:desc}; score: criteria [2-10 levels, low->high], answer is the 0-based level position (fractional allowed). Batch all questions per call. Out: {root?,results:{item:answer or {qid:answer}},unsure?}. noul -> P(yes) | [p,"?"]; choice/score -> [value,conf(,top2 probs)]. Open unsure items yourself. Weak at counting, math, dates, multi-step reasoning. |
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 1 tool
With only a single tool, there is no possibility of confusing one tool with another. The tool's multiple question types (noul, choice, score) are clearly delineated within the one tool, so agents can disambiguate the intended use from the parameter descriptions.
The sole tool name 'jev_ask' follows a clean snake_case convention with a descriptive verb. As there is no other tool to compare against, no naming inconsistencies or mixed conventions exist.
At one tool, the server is on the low end of tool counts, but for the narrow purpose of classifier-based file questioning, a single comprehensive tool is reasonable. It slightly under-utilizes the typical 3-15 tool range, yet it does not feel excessively thin given its focused scope.
The single tool covers multiple question types, batching, and uncertainty handling, addressing the full lifecycle of querying files without needing additional operations. The documented limitations (weak at counting, math, dates, multi-step reasoning) are clearly stated, so there are no hidden gaps in the tool surface.