impeccable
The design vocabulary for your coding agent. Steer with polish and bolder, review with critique and audit, catch AI slop before it ships.
Danbooru MCP server for AI image generation: cached search, batch lookup, profiles, wiki, posts
教 AI 做这类工作的 Skill,可与本项一起使用
The design vocabulary for your coding agent. Steer with polish and bolder, review with critique and audit, catch AI slop before it ships.
Editorial diagram design for Claude Code, Codex, GitHub Copilot, Cursor, Factory Droid, Qoder, and Pi. 45 diagram types. Self-contained HTML + SVG. No shadows. No Mermaid slop.
Create beautiful slides on the web using a coding agent's frontend skills
列表按使用分(满分 100)排序,由以下部分组成
| 热度 | 4/80 | GitHub 星标或每周下载量:较大的计 60,较小的计 20 |
|---|---|---|
| 新鲜度 | 12/12 | 30 天内有发布或代码更新得满分 |
| 仍在响应 | 0/8 | 上次检查时云端端点有响应 |
| 使用分 | 16/100 |
质量: 基础检查通过 3/3 项(仅展示,不用于排序) · 上升势头: 暂无数据 (只用于“上升中”标签)
选择你使用的 AI 应用,然后复制设置。
通过 3/3 项检查 · 在你的电脑上运行
第一步:你用哪个 AI?
选择你用的应用,步骤会显示在下方。
Claude 暂时无法连接 danbooru-tag-mcp
开发者尚未在 MCP 注册表中填写安装方式
ChatGPT 暂时无法连接 danbooru-tag-mcp
开发者尚未在 MCP 注册表中填写安装方式
B · 设计、图片与音频 类中能连接 ChatGPT 的其他服务器
从同类中使用较多的里挑选,仅限用地址即可连接、无需密钥的
Claude Code 暂时无法连接 danbooru-tag-mcp
开发者尚未在 MCP 注册表中填写安装方式
Cursor 暂时无法连接 danbooru-tag-mcp
开发者尚未在 MCP 注册表中填写安装方式
VS Code 暂时无法连接 danbooru-tag-mcp
开发者尚未在 MCP 注册表中填写安装方式
它在你的电脑上运行;安装后,工具列表会显示在你的 AI 应用中。
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月20日
注册名称已通过 GitHub 账户 BaixuanZhu 验证,该账户拥有此仓库
代码最后更新于8天前
使用 MIT
它在你的电脑上运行,没有互联网端点
在你的电脑上运行,改用本地设置
它在你的电脑上运行;我们只能读取互联网端点的工具列表
如果你是作者,可把这段代码贴到 README。徽章会链接回本页。
[](https://mcp.popnx.com/en/servers/baixuanzhu-danbooru-tag-mcp)提问、分享技巧或报告无法使用。按“有帮助”数量排序。
还没有评价
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
它在 3 项适用的基础检查中通过了 3 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
注册中心没有列出必需的密钥,不过有些服务器会在首次使用时要求你登录。
它在你的电脑上运行。你的 AI 应用会在需要时启动它,数据不会经过作者的服务器(除非它自己调用在线服务)。
GitHub 仓库最后更新于8天前,最新版本(0.6.0)于8天前进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。
同一分类中有:HyperFrames by HeyGen, Pascal 3D Editor, draw.io, Figma MCP Server, Remocn: Remotion MCP server
同一分类、其他作者、命令已通过检查。官方的排在前面,再按星标排序。
如何看卡片:↓ 每周下载量i★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
Official HyperFrames by HeyGen: the best AI video editor in chat. Captions, motion graphics, MP4.
Create, inspect, validate, and save editable 3D building scenes with Pascal's MCP server.
Create draw.io diagrams in chat from Mermaid or XML, shown live and interactive. 10,000+ shapes.
The Figma MCP server brings Figma design context directly into your AI workflow.
Hosted MCP for Remotion videos: search Remocn components, read docs, plan a video, preview it.
| 本项 danbooru-tag-mcp | HyperFrames by HeyGen | Pascal 3D Editor | draw.io | Figma MCP Server | Remocn: Remotion MCP server | |
|---|---|---|---|---|---|---|
| 运行位置 | 需要安装在电脑上 | 在云端运行,无需安装 | 云端或安装在电脑上都可以 | 在云端运行,无需安装 | 在云端运行,无需安装 | 在云端运行,无需安装 |
| 需要密钥 | 否 | 否 | 否 | 否 | 否 | 否 |
| 可在 Claude 中按 + 添加 | 否 | 是 | 否 | 否 | 是 | 否 |
| 基础检查 | 3/3 | 4/4 | 5/5 | 4/4 | 4/4 | 4/4 |
| 许可证 | MIT | Apache-2.0 | MIT | Apache-2.0 | – | MIT |
| 价格 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 |
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
English | 简体中文
A Danbooru tag lookup MCP server for local AI image generation. Your AI client looks tags up instead of recalling them — exact names, aliases, categories, post counts, co-occurrence stats and wiki rules, resolved before it writes the prompt.
Windows 10/11 on x64 or ARM64. Anonymous access works, so there is no API key to obtain before the first query.
Install per-user (Windows 10/11, no admin rights):
iwr -useb "https://raw.githubusercontent.com/BaixuanZhu/danbooru-tag-mcp/main/install.ps1" | iex
Point your MCP client at the command:
{
"mcpServers": { "danbooru-tags": { "command": "danbooru-tag-mcp", "args": [] } } }
That's it — the client picks up the eight tools below on its next session.
Live responses, abridged for length. Post counts move daily; these were captured on 2026-11-12.
search_tags for *blue hair* — the names that actually exist, so the model stops guessing bluehair:
{"tags": [
{"id": 10953, "name": "blue_hair", "category": 0, "post_count": 1234156}, {"id": 545704, "name": "blue_hairband", "category": 0, "post_count": 33332}, {"id": 1564040, "name": "blue_hair-chan_(ramchi)", "category": 4, "post_count": 120} ]}
get_tag_info for blue_hair — the count, plus whether the WD14 taggers know the name at all:
{"id": 10953, "name": "blue_hair", "category": 0, "post_count": 1234156,
"wd14": {"status": "live", "sources": [ {"vocab": "wd-eva02-large-tagger-v3", "count": 704952}, {"vocab": "wd-v1-4-moat-tagger-v2", "count": 564360}]}}
get_tag_profile for arm_up, all-ages sample of 200 — what the tag actually renders alongside, ranked by in-sample frequency, and which combinations are specific to it rather than merely common:
{"name": "arm_up", "post_count": 314588, "rating": "g", "sample": 200,
"ratings": {"g": 1}, "ubiquitous": [ {"tag": "smile", "category": 0, "freq": 0.51}, {"tag": "solo", "category": 0, "freq": 0.51}], "co_tags": [ {"tag": "looking_at_viewer", "category": 0, "freq": 0.5}, {"tag": "long_sleeves", "category": 0, "freq": 0.48}, {"tag": "1girl", "category": 0, "freq": 0.46}, {"tag": "open_mouth", "category": 0, "freq": 0.45}], "top_pairs": [ {"pair": "pokemon + pokemon_(creature)", "freq": 0.05, "lift": 16.7}, {"pair": "japanese_clothes + kimono", "freq": 0.05, "lift": 15.4}]}
The full profile returns 30 co-occurring tags and 10 ranked pairs. Every field is documented under Tools.
1girl/solo split into ubiquitous, combinations ranked by lift, and the sample's rating mix. with profiles a tag pair instead, answering what arm_up plus holding_weapon renders as.summary, intro_tags, rules, see_also and example_posts, so rules like "if X, use Y instead" arrive as data rather than prose to re-parse.live, tagger_era (renamed upstream, the canonical name included) or unknown, so LoRA card captions stop dead-ending at "tag not found".rating:g for all-ages only.upgrade verifying SHA256 and swapping the binary atomically, a per-user installer that registers PATH, and 1.5 s request spacing (500 ms with credentials) that keeps anonymous access viable.Tags come back as plain strings with no model coupling, so they paste straight into ComfyUI, Stable Diffusion or Illustrious prompt fields.
Requires Windows 10/11 on x64 or ARM64. Every option installs per-user, needs no admin rights, and puts the install directory on the user PATH, so the bare danbooru-tag-mcp command works afterwards.
Download danbooru-tag-mcp-windows- -setup.exe from the latest release and double-click it. It installs to %LOCALAPPDATA%\Programs\danbooru-tag-mcp.
iwr -useb "https://raw.githubusercontent.com/BaixuanZhu/danbooru-tag-mcp/main/install.ps1" | iex
Custom install directory:
.\install.ps1 -InstallDir "D:\tools\danbooru-tag-mcp"
Download danbooru-tag-mcp-windows- .zip from the release page, extract danbooru-tag-mcp.exe anywhere, and run it once — the first run silently adds its own directory to the user PATH (idempotent).
Point your MCP client (Claude Desktop, ZCode, etc.) at the command:
{
"mcpServers": { "danbooru-tags": { "command": "danbooru-tag-mcp", "args": [] } } }
The bare command name works because the install directory is on the user PATH; alternatively use the full exe path.
Add this repository as a plugin marketplace and install danbooru-tag-mcp — the plugin ships the danbooru-tags server definition, so there is no JSON to hand-edit. It wires configuration only: finish one of the install options above first, so the CLI is actually on PATH.
BaixuanZhu/danbooru-tag-mcp → install the danbooru-tag-mcp plugin./plugin install danbooru-tag-mcp --marketplace BaixuanZhu/danbooru-tag-mcp (one step, Claude Code 2.1.275+), or add the marketplace first with claude plugin marketplace add BaixuanZhu/danbooru-tag-mcp and install from there.danbooru-tag-mcp upgrade keeps the CLI updated independently of the plugin. See plugins/danbooru-tag-mcp/README.md.
| Tool | What it does |
|---|---|
search_tags | Search tags by keyword, ordered by post count |
get_tag_info | Exact info for one tag, with a WD14 verdict on the name |
get_tags_info | The same for up to 20 names in one request |
get_related_tags | Tags that co-occur with a given tag |
get_tag_alias | Resolve an alias or misspelling to its canonical tag |
get_tag_wiki | A tag's wiki page as structured fields |
search_posts | Posts matching a tag combination, with their tag lists |
get_tag_profile | What a tag actually renders as, from a post sample |
Parameters and return fields for all eight tools
search_tagsquery (required, e.g. blue hair) · limit (10) · category (general/artist/copyright/character/meta) · min_posts (0, drops tags below the threshold) · refresh (false)
When no result carries posts, the response adds wd14_hits — WD14 tagger vocabulary names containing the query.
get_tag_infoname (required, e.g. blue_hair) · refresh (false)
Returns category and post count plus a wd14 verdict: live (current Danbooru), tagger_era (WD14 vocabulary only, likely renamed — the canonical name is in renamed_to) or unknown (neither). A 0-post name with an active alias also returns renamed_to.
get_tags_infonames (required, e.g. ["blue_hair","1girl"], max 20) · refresh (false)
Returns found (each with the fields of get_tag_info, renamed_to included on 0-post rows) and missing (absent from Danbooru, each still carrying its wd14 verdict).
get_related_tagstag (required) · limit (10) · refresh (false)
Meta-category tags like highres are filtered out as noise.
get_tag_aliasname (required, e.g. sailor_suit) · refresh (false)
null means the input is already canonical.
get_tag_wikititle or other_names (one required) · limit (5) · include_description (false) · refresh (false)
Returns summary (first paragraph), intro_tags (definition-section link targets), rules (disambiguation and usage sentences), see_also, example_posts, the raw linked_tags union and multilingual other_names. The raw DText body is not returned; include_description adds the intro prose as plain text, for character appearance descriptions.
search_poststags (required, space-separated) · limit (5, max 20) · tags_mode (full default, none drops the per-post tag lists, diff keeps only tags not already in the query)
No rating filter by default; add a rating:g/s/q/e metatag to narrow. Each post carries score, fav_count, large_file_url and tags split by category: copyright / artist / character / general / meta.
get_tag_profiletag (required, e.g. arm_up) · sample (200, range 20-200) · with · rating (g/s/q/e) · order (random by default, or score / favcount for a top-post view, which is biased)
Samples a page of the tag's posts and returns co_tags ordered by in-sample frequency, ubiquitous corpus constants (freq >= 0.5), top_pairs combinations ranked by lift, and the ratings mix. Random sampling covers the whole population; rating filters the sample; with profiles the 200 most recent joint posts of a tag pair instead.
The six tag tools (search_tags, get_tag_info, get_tags_info, get_related_tags, get_tag_alias, get_tag_wiki) read from the response cache: repeat calls inside the TTL skip the network round trip, and during a Danbooru outage the last cached answer is served instead of an error, marked with stale: true and cached_at (when it was stored). Their refresh parameter (default false) bypasses the cache for a single call and re-fetches live. Post search and profiles are never cached, so they carry no refresh.
Repeat queries are answered from an on-disk cache instead of hitting Danbooru again:
| Data | TTL |
|---|---|
Related tags (get_related_tags) | 30 days |
| Wiki pages, aliases | 7 days |
Tag searches (search_tags) | 24 hours |
Exact tag lookups (get_tag_info, get_tags_info) | 7 days |
Posts (search_posts, get_tag_profile) | never cached |
%LOCALAPPDATA%\danbooru-tag-mcp\cache (per user, persists across restarts, capped at 2000 entries / 64 MB with oldest-first eviction). It holds only API response bodies and is safe to delete at any time; the uninstaller removes it.stale: true plus cached_at (when it was stored) on the affe