stock-image-mcp
An MCP server for searching and downloading stock images from Unsplash, Pexels, Pixabay, Freepik, StockVault, and Burst, enabling AI agents to source images for content creation via tool calls.
README
stock-image-mcp
<!-- mcp-name: io.github.Hasilt/images-mcp -->
An MCP server that lets Claude Code (or any MCP-compatible agent) search and download stock images by tool call — useful for sourcing images while writing SEO blog posts or other content.
Providers
| Provider | API type | Requires key | Default rate limit |
|---|---|---|---|
| Unsplash | Official | Yes | 50/hr demo, 5000/hr production |
| Pexels | Official | Yes | 200/hr |
| Pixabay | Official | Yes | 100 req/60s |
| Freepik | Official | Yes | configurable (plan-dependent) |
| StockVault | Official | Yes | configurable (undocumented, conservative default) |
| Burst (Shopify) | Unofficial scrape | No | self-imposed, polite default |
Burst has no public API. It's included as a best-effort HTML scraper, clearly
marked unsupported in code — any failure there is caught and reported as an
empty result rather than breaking search_all_images.
Install
Each user runs the server locally and supplies their own provider API keys — there's no shared hosting or centrally-held keys.
With Claude Code:
claude mcp add stock-image-mcp \
-e UNSPLASH_ACCESS_KEY=... \
-e UNSPLASH_TIER=demo \
-e PEXELS_API_KEY=... \
-e PIXABAY_API_KEY=... \
-e FREEPIK_API_KEY=... \
-e FREEPIK_REQUESTS_PER_MINUTE=60 \
-e STOCKVAULT_API_KEY=... \
-e STOCKVAULT_REQUESTS_PER_HOUR=60 \
-e BURST_REQUESTS_PER_MINUTE=10 \
-e DEFAULT_PROVIDER=pexels \
-e DOWNLOAD_DIR=./downloads \
-- uvx stock-image-mcp
With OpenAI Codex CLI:
codex mcp add stock-image-mcp \
--env UNSPLASH_ACCESS_KEY=... \
--env UNSPLASH_TIER=demo \
--env PEXELS_API_KEY=... \
--env PIXABAY_API_KEY=... \
--env FREEPIK_API_KEY=... \
--env FREEPIK_REQUESTS_PER_MINUTE=60 \
--env STOCKVAULT_API_KEY=... \
--env STOCKVAULT_REQUESTS_PER_HOUR=60 \
--env BURST_REQUESTS_PER_MINUTE=10 \
--env DEFAULT_PROVIDER=pexels \
--env DOWNLOAD_DIR=./downloads \
-- uvx stock-image-mcp
With Gemini CLI:
gemini mcp add stock-image-mcp \
-e UNSPLASH_ACCESS_KEY=... \
-e UNSPLASH_TIER=demo \
-e PEXELS_API_KEY=... \
-e PIXABAY_API_KEY=... \
-e FREEPIK_API_KEY=... \
-e FREEPIK_REQUESTS_PER_MINUTE=60 \
-e STOCKVAULT_API_KEY=... \
-e STOCKVAULT_REQUESTS_PER_HOUR=60 \
-e BURST_REQUESTS_PER_MINUTE=10 \
-e DEFAULT_PROVIDER=pexels \
-e DOWNLOAD_DIR=./downloads \
-- uvx stock-image-mcp
Or add it to your MCP config manually with uvx
(no clone or install step required — uvx fetches the package from PyPI on
first run):
{
"mcpServers": {
"stock-image-mcp": {
"command": "uvx",
"args": ["stock-image-mcp"],
"env": {
"UNSPLASH_ACCESS_KEY": "...",
"UNSPLASH_TIER": "demo",
"PEXELS_API_KEY": "...",
"PIXABAY_API_KEY": "...",
"FREEPIK_API_KEY": "...",
"FREEPIK_REQUESTS_PER_MINUTE": "60",
"STOCKVAULT_API_KEY": "...",
"STOCKVAULT_REQUESTS_PER_HOUR": "60",
"BURST_REQUESTS_PER_MINUTE": "10",
"DEFAULT_PROVIDER": "pexels",
"DOWNLOAD_DIR": "./downloads"
}
}
}
}
Only UNSPLASH_ACCESS_KEY, PEXELS_API_KEY, PIXABAY_API_KEY,
FREEPIK_API_KEY, and STOCKVAULT_API_KEY are actual secrets — set only the
ones for providers you want enabled and drop the rest; any provider key you
omit is simply skipped by search_all_images and rejected if queried
directly via search_stock_images. The remaining variables are optional
tuning knobs shown above with their defaults — see .env.example for the
full list.
Tools
search_stock_images(query, provider="default", orientation=None, per_page=10, page=1)search_all_images(query, orientation=None, per_page=5)— fans out to every configured provider concurrentlyget_best_image(query, provider="default")download_image(url, dest_path, provider=None)— saves locally, returns attribution text if the image came from a prior searchget_attribution(provider, image_id)get_rate_limit_status(provider=None)
Usage
You don't call these tools directly — you just talk to your agent, and it decides when to reach for one based on what you asked and the tool descriptions above. A few things worth knowing:
- Just ask in plain English. "Find me 3 landscape photos of mountains for a
blog post" is enough to trigger
search_all_images(orsearch_stock_imagesif you name a provider). "Download that first Unsplash result to./images/hero.jpg" triggersdownload_image. "How many Pexels requests do I have left this hour?" triggersget_rate_limit_status. search_stock_imagesneeds a specific provider (or"default");search_all_imagesjust queries everything you've configured keys for and merges the results. Say "search Pexels only" or similar if you care which one gets used.- Several providers (Unsplash and Pexels in particular) require attribution
if you actually use the image somewhere public.
download_imagereturns the attribution text alongside the file when it applies — it's on you to paste it wherever the image ends up, the server won't do that part for you. - Want to call tools by hand instead of through a conversation? See the MCP Inspector section under Development below.
Developing locally
Working from a clone instead of the published package:
uv sync
cp .env.example .env # fill in the API keys for providers you want enabled
uv run stock-image-mcp
Point your Claude config at the local checkout instead of uvx:
{
"mcpServers": {
"stock-image-mcp": {
"command": "uv",
"args": [
"--directory",
"/absolute/path/to/stock-image-mcp",
"run",
"stock-image-mcp"
],
"env": {
"UNSPLASH_ACCESS_KEY": "...",
"UNSPLASH_TIER": "demo",
"PEXELS_API_KEY": "...",
"PIXABAY_API_KEY": "...",
"FREEPIK_API_KEY": "...",
"FREEPIK_REQUESTS_PER_MINUTE": "60",
"STOCKVAULT_API_KEY": "...",
"STOCKVAULT_REQUESTS_PER_HOUR": "60",
"BURST_REQUESTS_PER_MINUTE": "10",
"DEFAULT_PROVIDER": "pexels",
"DOWNLOAD_DIR": "./downloads"
}
}
}
}
Development
uv run pytest # test suite (mocked HTTP, no live keys needed)
uv run ruff check . # lint
uv run ruff format . # format
uv run mypy src # type check
To try it against real providers, use the MCP Inspector:
npx @modelcontextprotocol/inspector uv run stock-image-mcp
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。