k-exaone-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| EXAONE_MODEL | No | 사용할 모델 ID. | k-exaone_v2.5 |
| EXAONE_TOP_P | No | 누적 확률 상한(공식 권장값 0.95). 0 ~ 1. | 0.95 |
| EXAONE_API_KEY | Yes | EXAONE API 키 (필수). EXAONE API Platform에서 발급받습니다. | |
| EXAONE_WORKDIR | No | files에 상대 경로가 들어왔을 때 기준이 되는 폴더. 기본값은 서버 실행 폴더입니다. | |
| EXAONE_BASE_URL | No | OpenAI 호환으로 띄운 다른 EXAONE 엔드포인트(vLLM 등)에 연결할 때 바꿉니다. | https://api.exaone.ai/v1 |
| EXAONE_MAX_TOKENS | No | 기본 생성 토큰 상한(추론 포함). 1 이상. 비워 두면 API 서버 기본값을 씁니다. | |
| EXAONE_TIMEOUT_MS | No | EXAONE 응답을 기다리는 최대 시간(ms). 1000 이상. 예: 120000=2분, 300000=5분, 1200000=20분. | 600000 |
| EXAONE_MAX_FILE_KB | No | 첨부 파일 하나의 최대 크기(KB). 1 이상. | 512 |
| EXAONE_TEMPERATURE | No | 샘플링 무작위성(공식 권장값 1.0). 0 이상. | 1.0 |
| EXAONE_MAX_TOTAL_KB | No | 한 번에 첨부하는 파일 합계의 최대 크기(KB). 1 이상. | 1536 |
| EXAONE_REASONING_EFFORT | No | 호출 시 reasoning_effort를 지정하지 않았을 때 사용할 기본 추론 강도. 허용 값: low, medium, high, max. 낮을수록 빠르고 토큰을 적게 씁니다. | high |
| EXAONE_PRESERVE_THINKING | No | 대화를 이어 갈 때 이전 추론을 함께 보낼지 여부. 허용 값: true, false. false면 입력 토큰이 줄어듭니다. | true |
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 |
|---|---|
| ask_exaoneA | Delegate a task or question to K-EXAONE 2.5 (LG AI Research) and return its answer. Use it to get an independent second opinion, a Korean-native draft or review, or to offload a self-contained subtask.
EXAONE cannot see your workspace, tools or this conversation: put every needed fact in |
| exaone_statusA | Show k-exaone-mcp settings, token usage accumulated by this server process, and the latest EXAONE API rate-limit headers. With ping=true (default) it also sends one tiny request to verify the API key and connectivity. |
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 2 tools
The two tools occupy completely separate roles: ask_exaone performs model delegation/inference, while exaone_status reports configuration, token usage, and API health. An agent cannot plausibly confuse a query tool with a diagnostics tool, and the descriptions reinforce the boundary.
Both names are snake_case and share the `exaone` token, giving a recognisable family. However, ask_exaone is verb-first while exaone_status is prefix-first, a minor structural deviation rather than a real inconsistency.
Two tools is thin in absolute terms, but the server's scope is deliberately narrow (single-model delegation plus health check), and both tools earn their place. Nothing is missing that would warrant an extra tool purely for count's sake.
The core lifecycle is covered: stateless asks, file-passing, multi-turn follow-ups via conversation_id, and a status/connectivity check. Gaps are minor — there is no way to list, inspect, or delete existing conversation threads, and no parameter/model discovery beyond status.