Skip to main content
Glama

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault
GAMETRANS_MODELNoModel name to use for translation, e.g. deepseek-chat
GAMETRANS_API_KEYNoTranslation API key for an OpenAI-compatible service.
GAMETRANS_BASE_URLNoBase URL for the OpenAI-compatible translation API, e.g. https://api.deepseek.com/v1

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

CapabilityDetails
tools
{
  "listChanged": false
}

Tools

Functions exposed to the LLM to take actions

NameDescription
project_initB

初始化 gametrans 工作区(自动探测引擎)

project_statusC

查看项目各层的当前状态

engine_listC

列出已安装的引擎支持包

engine_infoC

查看某个引擎支持包的能力

engine_detectC

探测目录使用的是什么引擎

engine_optionsB

查看当前引擎的私有选项与外部工具链(官方 SDK)状态

engine_option_setB

设置一条引擎私有选项(例如外部工具链的路径)

engine_option_clearC

删掉一条引擎私有选项

scanC

提取全部待译内容,产出带权路径图

graph_showC

按权重列出路径图节点

graph_nodeC

按逻辑路径/node_id/unit_id 查看一个节点

graph_dependenciesB

列出依赖边:候选(方向未确认)与已确认分开报

graph_dependB

记下/确认一条依赖边(确认方向后它才参与调度排序)

graph_undependC

删掉一条依赖边

graph_knowledgeB

把知识边写进图(术语书 + 原文命中算出来的先后),或撤掉

irB

导出 Localization IR 并校验提取层一致性

skeleton_statusB

查看引擎生成的骨架里有什么(槽位数、两种定键方式的分布、按文件的拆分)

language_factsA

查看语言包事实:游戏认哪些语言代码、字体从哪来、语言目录里现有什么

translateC

按路径图翻译(串行或并行)

planB

看这次翻译会按什么顺序跑:区域 → 有序阶段(可指定调度策略)

stalenessA

点名该重做的译文:过期 / 缺失 / 不可用分开报(知识改了走这条精确回补)

writebackC

把译文写回游戏,生成翻译层(默认只填空缺,不动产物里已有的译文)

revalidateA

按当前判据重新裁定当初被挡下的译文(只放行,不收紧)

unifyA

统一替换:把已落盘译文里撞车过的旧写法换成那一条的定译(裁决之后收口用;替换源只来自撞车证据,逐条重过结构校验并留痕)

packC

把翻译产物封包成可分发补丁

tasksA

查看持久化的翻译任务状态;给 --ref 则看某一条的完整载荷(含检索结果)

agent_nextC

取最老的一条挂单并认领(本该发给 API 的完整请求,由 agent 用自己的额度作答)

agent_submitB

交一份答案给挂单(只有 pending 的单能收;答案回到同一条校验链)

agent_statusB

看挂单队列现状:几条在等、几条已答未收

resource_term_listA

列出术语书(一行一个实体:key 写法列表 / profile 事实列表 / constant / order / position)

resource_term_addC

人拍的板:整行按给进来的那份写(五栏都可以只填一栏;找不到同一行就新开一行)

resource_term_removeC

删除一行(按任一写法)

resource_term_importB

整份换上:让术语书最终恰好是这份 JSONL 里的几行 —— 拆行 / 合行 / 批量改名一次做完,每一步都进变更日志

resource_term_proposeA

提一条更正(改已有的译名 / 某条事实):先排队,采用之前不生效

resource_term_pending_listA

列出待审更正(改已有的译名 / 事实)——采用之前书里一个字节都不动

resource_term_pending_adoptA

采用一条待审更正:把 old 换成 new、记变更日志、从队列里删掉(模型身份会被拒)

resource_term_pending_dropA

丢弃一条待审更正(不改书,只从队列里删掉)

resource_validateB

校验资源文件,列出非法行

resource_term_candidatesA

从场摘要抽实体(一个实体一行,新写法 / 空译名 / 新事实免审追加,改已有的译名进待审队列;幂等,重复调用不重复产行)

resource_harvestB

把游戏里已有的译文读进来,沉淀成翻译记忆(只读游戏;观察出来的对应不进术语书)

resource_deviation_listA

列出声明的偏离(有意留空 / 经批准的表达式改写)

resource_deviation_addA

声明一条偏离:这条译文有意偏离判据的哪一条(默认不许,声明才放行)

resource_deviation_removeC

删掉一条偏离声明

resource_supplement_listC

列出声明的补充条目(引擎不枚举、但玩家看得见的文本)

resource_supplement_addA

声明一条补充条目:这段文本也要翻(说话人名、代码里的界面提示等)

resource_supplement_removeC

删掉一条补充条目

resource_style_listC

列出风格要求(一等翻译资源)

resource_style_addB

新增/覆盖一条风格要求(按 aspect + scope 定键)

resource_style_removeC

删掉一条风格要求

ui_viewsB

查看交互层视图(默认只列用户可见的)

ui_renderC

按可见性策略渲染给用户的文本

ui_policy_setC

设置某类信息对用户可见还是透明

ui_announceB

以 agent 身份直接向用户发布一条信息

ui_overrideC

改写一条交互层视图的内容

ui_hideB

对用户隐藏(或恢复)一条视图

ui_clearB

清掉非置顶的视图

config_showA

查看当前配置(含每一项来自哪一层)

config_pathC

报出四层配置各自的文件在哪

config_setB

修改一项配置;默认改项目层,--global 改全局层(一次配好、所有项目通用)

config_unsetA

删掉某一层的某一项,让它回落下一层(不是设成默认值)

config_providersC

列出可用的翻译 provider

Prompts

Interactive templates invoked by user choice

NameDescription

No prompts

Resources

Contextual data attached and managed by the client

NameDescription

No resources

TDQS

C2.9/5.0

Scored across 61 tools

Disambiguation4/5

Most tools have clearly distinct purposes and the detailed Chinese descriptions help distinguish operations. A few pairs like engine_options/engine_info, resource_term_add/resource_term_import, and graph_show/graph_dependencies have some overlap, but boundaries are generally discoverable.

Naming Consistency4/5

All names use snake_case, which is consistent. However, the set mixes prefixed tools (engine_, graph_, resource_) with bare verbs/nouns (pack, scan, translate, plan, staleness), and mixes verb-first with noun-first patterns.

Tool Count1/5

61 tools is an extreme mismatch for practical MCP scoping. Even for a complex localization pipeline, this imposes a very high selection and maintenance burden, far beyond the recommended 3-15 range.

Completeness4/5

The surface covers project setup, engine discovery, graph/dependency management, translation lifecycle, resource CRUD, UI/config, and agent queue operations. Minor gaps remain, such as project deletion or engine package lifecycle management, but core workflows are well covered.

Maintenance

ActivityMaintained
ResponsivenessNo issues