Gelbooru MCP
A Python MCP server that wraps the Gelbooru API. Connect it to any MCP-compatible client (Claude Desktop, Cursor, etc.) to search posts, look up tags, and generate Stable Diffusion prompts from real character appearance data — all directly from your AI assistant.
README
gelbooru-mcp
A Python MCP server that wraps the Gelbooru API. Connect it to any MCP-compatible client (Claude Desktop, Cursor, etc.) to search posts, look up tags, and generate Stable Diffusion prompts from real character appearance data — all directly from your AI assistant.
✨ Features
🎨 Stable Diffusion Prompt Generation
- Character Prompts: Auto-generate accurate SD prompts from real Gelbooru tag frequency data
- Appearance Breakdown: Separate eye, hair, and clothing/accessory tag categories
- Smart Caching: Results cached for 24 hours — no repeated API hits
🔍 Post & Tag Search
- Advanced Filtering: Search by tags, score, resolution, rating, uploader, pool, and more
- Full Tag Syntax: AND, OR, wildcard, exclusion, meta-tags, sorting, and pagination
- Tag Lookup: Check tag existence, post counts, and discover related tags
👥 Community Tools
- User Search: Find Gelbooru user accounts by name or wildcard pattern
- Comments: Retrieve post comments for any post ID
- Deleted Posts: Track removed content above a given post ID
📦 Installation
Prerequisites
- Python 3.10+
git
Quick Start
- Clone the repository:
git clone https://github.com/citronlegacy/gelbooru-mcp.git
cd gelbooru-mcp
- Run the installer:
chmod +x install.sh && ./install.sh
# or without chmod:
bash install.sh
- Or install manually:
pip install mcp
Note: Add
.gelbooru_cache/and.venv/to your.gitignoreto avoid committing cached data or your virtual environment.
Getting a Gelbooru API Key
- Visit your Gelbooru account options page
- Log in to your Gelbooru account
- Copy your API Key and User ID
- Set them as environment variables (see below)
🔑 Authentication
API credentials are optional but strongly recommended — unauthenticated requests are throttled and limited to 2 tags per query. Gelbooru Patreon supporters receive unlimited requests.
export GELBOORU_API_KEY="your_api_key"
export GELBOORU_USER_ID="your_user_id"
Both values are on your Gelbooru account options page. Without them the server still works but requests may be throttled. Patreon supporters of Gelbooru are not rate-limited.
▶️ Running the Server
python gelbooru_mcp.py
# or via the venv created by install.sh:
.venv/bin/python gelbooru_mcp.py
⚙️ Configuration
Claude Desktop
Add the following to your claude_desktop_config.json:
{
"mcpServers": {
"gelbooru-mcp": {
"command": "/absolute/path/to/.venv/bin/python",
"args": ["/absolute/path/to/gelbooru_mcp.py"],
"env": {
"GELBOORU_API_KEY": "your_api_key",
"GELBOORU_USER_ID": "your_user_id"
}
}
}
}
Other MCP Clients
Configure according to your client's documentation:
- Command:
/absolute/path/to/.venv/bin/python - Args:
/absolute/path/to/gelbooru_mcp.py - Transport: stdio
💡 Usage Examples
Generate a Stable Diffusion prompt for a character
"Build me a Stable Diffusion prompt for Rem from Re:Zero."
The LLM calls build_prompt with character_name: "rem_(re:zero)" and gets back:
rem (re:zero), blue eyes, blue hair, short hair, maid, maid headdress, maid apron, ...
Find high-quality wallpaper images
"Show me the top-rated scenery images that are at least 1920px wide."
The LLM calls search_posts with tags: "scenery width:>=1920 sort:score:desc".
Look up how popular a tag is
"How many posts does the tag 'misty_(pokemon)' have on Gelbooru?"
The LLM calls search_tags with name: "misty_(pokemon)" and reads the count field.
🛠️ Available Tools
| Tool | Description | Key Parameters |
|---|---|---|
build_prompt |
Generate a Stable Diffusion prompt string for a character | character_name, max_images, include_other |
get_character_tags |
Get structured tag breakdown with frequency counts | character_name, max_images |
search_posts |
Search posts with full tag syntax support | tags, limit, pid, id |
search_tags |
Look up tags by name, pattern, or ID | name, name_pattern, orderby, limit |
search_users |
Find Gelbooru user accounts | name, name_pattern, limit |
get_comments |
Retrieve comments for a post | post_id |
get_deleted_posts |
List recently deleted posts | last_id, limit |
📖 Tools Reference
build_prompt
Fetches the most-tagged rating:general solo posts for a character and assembles a ready-to-paste Stable Diffusion prompt string. Internally calls get_character_tags so results are cached after the first fetch.
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
character_name |
string | ✅ | — | Gelbooru character tag, e.g. misty_(pokemon) |
max_images |
integer | ❌ | 300 |
Posts to analyse. More = slower first fetch, more reliable tags. Cached afterward. |
include_other |
boolean | ❌ | true |
Include non-eye/hair tags (clothing, accessories, etc.). Set to false for appearance-only prompts. |
Example response
{
"prompt": {
"character": "misty (pokemon)",
"posts_analysed": 284,
"cache_hit": false,
"prompt_string": "misty (pokemon), green eyes, orange hair, side ponytail, gym leader, shorts, suspenders",
"tags": {
"eye": ["green eyes"],
"hair": ["orange hair", "side ponytail"],
"other": ["gym leader", "shorts", "suspenders"]
}
}
}
LLM Tip: Always use Gelbooru's underscore format (
misty_(pokemon)notMisty (Pokemon)). If unsure of the exact tag, callsearch_tagswithname_patternfirst.
get_character_tags
Same data source as build_prompt but returns the full structured tag breakdown with frequency counts. Use this when you want to inspect, filter, or reformat tags yourself.
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
character_name |
string | ✅ | — | Gelbooru character tag, e.g. rem_(re:zero) |
max_images |
integer | ❌ | 300 |
Number of top-scored posts to analyse. |
Example response
{
"character_tags": {
"name": "rem_(re:zero)",
"posts_analysed": 300,
"cache_hit": true,
"eye": [
{ "tag": "blue eyes", "count": 261, "frequency": 0.87 }
],
"hair": [
{ "tag": "blue hair", "count": 274, "frequency": 0.913 },
{ "tag": "short hair", "count": 198, "frequency": 0.66 }
],
"other": [
{ "tag": "maid", "count": 231, "frequency": 0.77 }
]
}
}
frequency is the fraction of analysed posts that had that tag (0.0–1.0). Tags near 1.0 are near-universal; tags below 0.3 are situational.
Cache environment variables:
| Env var | Default | Description |
|---|---|---|
GELBOORU_CACHE_DIR |
.gelbooru_cache/ next to script |
Custom cache folder path |
GELBOORU_CACHE_TTL |
86400 (24 hours) |
Cache lifetime in seconds |
search_posts
Search Gelbooru posts using any tag combination. Supports the full Gelbooru tag syntax including meta-tags, sorting, and filtering.
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
tags |
string | ❌ | — | Tag query string (see Tag Syntax Reference below) |
limit |
integer | ❌ | 20 |
Posts to return (max 100) |
pid |
integer | ❌ | 0 |
Page number (0-indexed) for pagination |
id |
integer | ❌ | — | Fetch a single post by its Gelbooru post ID |
cid |
integer | ❌ | — | Fetch posts by change ID (Unix timestamp) |
get_deleted_posts
Retrieve posts that have been deleted from Gelbooru, optionally filtered to those above a given post ID.
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
last_id |
integer | ❌ | — | Only return deleted posts whose ID is above this value. Useful for syncing a local mirror. |
limit |
integer | ❌ | 20 |
Posts to return (max 100) |
search_tags
Look up Gelbooru tags by name, wildcard pattern, or ID. Useful for checking tag existence, finding post counts, discovering related tags, or autocomplete.
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
name |
string | ❌ | — | Exact tag name, e.g. blue_hair |
names |
string | ❌ | — | Space-separated list of exact tag names, e.g. cat_ears dog_ears fox_ears |
name_pattern |
string | ❌ | — | SQL LIKE wildcard: % = any chars, _ = one char. E.g. %schoolgirl% |
id |
integer | ❌ | — | Look up a tag by its database ID |
after_id |
integer | ❌ | — | Return tags with ID greater than this value |
limit |
integer | ❌ | 20 |
Tags to return (max 100) |
order |
string | ❌ | — | ASC or DESC |
orderby |
string | ❌ | — | Sort field: date, count, or name |
LLM Tip: If the user gives a character name in natural language (e.g. "Misty from Pokemon"), use
name_patternwith%misty%_(pokemon)%to find the correct Gelbooru tag before callingget_character_tagsorbuild_prompt.
search_users
Search for Gelbooru user accounts by name.
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
name |
string | ❌ | — | Exact username |
name_pattern |
string | ❌ | — | SQL LIKE wildcard username search |
limit |
integer | ❌ | 20 |
Results to return (max 100) |
pid |
integer | ❌ | 0 |
Page number for pagination |
get_comments
Retrieve comments on a specific post.
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
post_id |
integer | ✅ | — | The Gelbooru post ID to fetch comments for |
🏷️ Tag Syntax Reference
| Syntax | Meaning |
|---|---|
tag1 tag2 |
Posts with both tag1 AND tag2 |
{tag1 ~ tag2 ~ tag3} |
Posts with tag1 OR tag2 OR tag3 |
-tag1 |
Exclude posts with tag1 |
*tag1 |
Wildcard prefix (tags ending with tag1) |
tag1* |
Wildcard suffix (tags starting with tag1) |
rating:general |
Filter by rating: general, questionable, or explicit |
-rating:explicit |
Exclude a rating |
score:>=50 |
Score at least 50 |
width:>=1920 |
Image width ≥ 1920px |
height:>1080 |
Image height > 1080px |
user:bob |
Uploaded by user "bob" |
fav:1 |
Posts favourited by user with ID 1 |
pool:2 |
Posts in pool ID 2 |
sort:score:desc |
Sort by score (desc or asc) |
sort:random |
Random order on every request |
sort:random:1234 |
Random order with a fixed seed (0–10000) |
sort:updated:desc |
Sort by most recently updated |
🤖 Notes for LLMs
- Character tag format: Gelbooru uses
character_(series)format with underscores. Always convert natural language names before passing to tools — "Rem from Re:Zero" →rem_(re:zero), "Saber from Fate" →saber_(fate). - Workflow for character prompts: If the exact tag is unknown, call
search_tagswithname_patternfirst → confirm the tag exists and has posts → then callbuild_prompt. - Pagination:
search_postsreturns max 100 results per call. Usepidto walk through pages.get_character_tagsandbuild_prompthandle their own pagination internally. - Cache:
get_character_tagsandbuild_promptcache results for 24 hours. Thecache_hitfield in the response indicates whether live or cached data was used. - Ratings: Gelbooru uses
general,questionable, andexplicit.get_character_tagsandbuild_promptalways filter torating:generalfor cleaner, more representative character data.
⚠️ Known Limitations
- Tag search limit: Gelbooru enforces a maximum of 2 tags per unauthenticated search query. For complex multi-tag queries, set
GELBOORU_API_KEYandGELBOORU_USER_ID. get_character_tagsaccuracy: Results depend on how consistently a character is tagged on Gelbooru. Niche or recently added characters may have fewer posts and less reliable frequency data.rating:generalonly for character tools:build_promptandget_character_tagsintentionally restrict torating:generalfor clean, representative appearance data. Explicit posts are excluded by design.- Cache is per
(character_name, max_images)pair: Changingmax_imagesbusts the cache for that character.
🐛 Troubleshooting
Server won't start:
- Ensure Python 3.10+ is installed:
python --version - Verify the virtual environment was created:
ls .venv/ - Re-run the installer:
bash install.sh
API rate limiting / throttled requests:
- Set
GELBOORU_API_KEYandGELBOORU_USER_IDenvironment variables - Gelbooru Patreon supporters receive unlimited requests
Character not found / empty results:
- Confirm the tag exists with
search_tagsusingname_pattern - Check spelling — Gelbooru uses
character_(series)underscore format - Some characters may have very few consistently tagged posts
Tag syntax errors / too many tags:
- Unauthenticated users are limited to 2 tags per query
- Authenticate with API credentials for complex multi-tag searches
🤝 Contributing
Pull requests are welcome! If you find a character tag being miscategorised (e.g. a hair style tag missing from the hair bucket, or a noise tag slipping through the purge filter), please open an issue or PR with the tag and which list it belongs in.
Development Setup
- Fork the repository
- Create a feature branch
- Make your changes
- Submit a pull request
📄 License
MIT License — see LICENSE for details.
🔗 Links
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.