mcp_jev
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| JEV_MODEL | No | Model to use for Jev. | jev-latest |
| MCP_JEV_HOME | No | Config dir that stores the key and wrapper. | ~/.mcp_jev |
| MCP_JEV_CONFIG | No | Alias for MCP_JEV_HOME. | |
| MCP_JEV_CHECKOUT | No | Git checkout path (default ~/ mcp_jev or the repo you ran the script from). | |
| TYPESAFE_API_KEY | Yes | TypeSafe API key. Required for run_pack. Normally configured in ~/.mcp_jev/.env; process env is a fallback only when the store is empty. | |
| TYPESAFE_BASE_URL | No | Base URL for the TypeSafe API. | https://api.typesafe.ai |
| TYPESAFE_DEFAULT_MODEL | No | Used if JEV_MODEL is unset. |
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 |
|---|---|
| list_packsA | List the closed catalog of TypeSafe Jev packs. Packs are shortcuts: if an id fits, describe_pack + run_pack. If no pack fits, do not stop — use run_questions with closed state + typed Choice/Noul/Score. There is no free-form ask tool. |
| describe_packA | Return a pack's JSON Schema state, Choice/Noul/Score questions, example state, and suggested agent workflow. Always describe before run_pack when the schema is unfamiliar. |
| run_packA | Validate state and call TypeSafe System One (Jev) for that pack only. Returns typed answers (choice/noul/score), probabilities/confidence where the API provides them, and usage. Requires TYPESAFE_API_KEY. Does not run side effects. |
| run_questionsA | Typed custom path when no pack shortcut fits. Validate caller-built state plus typed Choice / Noul / Score and call the same TypeSafe System One path as run_pack. Not chat and not essay generation. Prefer run_pack when list_packs has an id (e.g. i18n_copy). Fail-closed schema: Choice is a closed option map (max 255; include a refuse key if needed), Noul is yes/no, Score is ≥2 ordered levels. Returns typed answers + usage. Requires TYPESAFE_API_KEY. No side effects. |
| pingA | Health: server version, TypeSafe JS SDK version, pack count, whether TYPESAFE_API_KEY is set. Never echoes the key. |
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 has a distinct role: listing packs, describing a pack, running a pack, running custom questions, and health checking. The only potential confusion is between run_pack and run_questions since both invoke the same TypeSafe System One path, but their descriptions clearly separate pack-based from custom typed workflows.
Most tools follow a clear verb_noun pattern: list_packs, run_questions, describe_pack, run_pack. The one outlier is the one-word health tool 'ping', which breaks the pattern but is a common and recognizable convention.
Five tools is a well-scoped size for a focused server that handles pack listing, description, execution, custom question runs, and health checks. Each tool has a clear purpose and none feel redundant.
The server covers the full intended workflow: discover packs via list_packs, understand via describe_pack, execute via run_pack, handle non-pack cases via run_questions, and verify health via ping. Since the pack catalog is closed and free-form ask is intentionally excluded, there are no significant gaps.