334k+ icons, inside your AI agent.
Pricing
Every plan: all 4 tools, all 334,395 icons, fair-use 60 requests/min, no hard daily cap.
Lifetime
- Pay once, never billed again
- One key, unlimited machines
- Lifetime updates + email support
Subscription
or $79/year
Same access, billed monthly or yearly. Replaces the lifetime deal when it ends.
What you get
Four tools your agent calls mid-task. You keep coding.
search_icons Search by intent
Full-text search across 334k+ icons. "shopping cart outline" → the right icon, first try.
search_icons(query: string, limit?: number) → { pack, name, svgUrl }[] get_icon Fetch the SVG
Raw SVG by pack + name, ready to paste into the file your agent is editing.
get_icon(pack: string, name: string) → { svg: string } list_collections Browse all packs
Every pack with license, count, and author. 236 collections.
list_collections() → { prefix, name, license, total }[] send_feedback Talk to us
Feedback straight from the agent session. No context switch.
send_feedback(message: string) → { ok: boolean } allsvgicons MCP vs. building it yourself
What an agent has to do either way.
| Need | allsvgicons MCP | DIY / raw API |
|---|---|---|
| Find the right icon for a description | search_icons does full-text intent matching in one call | Agent must fetch a pack index, then guess a filename |
| Get usable SVG markup | get_icon returns ready-to-paste SVG | Agent parses a sprite or icon-font mapping by hand |
| Know which pack/license to use | list_collections returns license + count per pack | Agent cross-references separate npm packages and READMEs |
| Rate limits | 60 req/min fair-use, headers show remaining headroom | Hard daily caps or no limit visibility, agent loops blind |
| Setup | One MCP config block, works over HTTP transport | Install an SDK, write a wrapper tool, maintain it |
Setup
One command once your key arrives, or hand the whole thing to your agent.
Fastest path
Copy this, paste it into your agent's chat. It configures itself.
Set up the allsvgicons MCP server in this project.
1. Add this MCP server config (for Claude Code, run: claude mcp add allsvgicons --transport http https://i.allsvgicons.com/mcp --header "Authorization: Bearer YOUR_API_KEY". For other clients, add this JSON to the MCP config file):
{
"mcpServers": {
"allsvgicons": {
"url": "https://i.allsvgicons.com/mcp",
"headers": { "Authorization": "Bearer YOUR_API_KEY" }
}
}
}
2. Replace YOUR_API_KEY with my real key.
3. Confirm the connection by calling list_collections.
4. From now on, when I ask for an icon: call search_icons with my description, then get_icon to fetch the SVG, and insert it directly into the file I'm editing. Claude Code
claude mcp add allsvgicons --transport http https://i.allsvgicons.com/mcp --header "Authorization: Bearer YOUR_API_KEY"Cursor / Generic JSON
{
"mcpServers": {
"allsvgicons": {
"url": "https://i.allsvgicons.com/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}FAQ
What is MCP?
MCP (Model Context Protocol) is an open standard that lets AI agents like Claude Code, Claude Desktop, Cursor, and Windsurf call external tools during a conversation. Our MCP server exposes icon search and retrieval as tools your agent can call directly.
Which clients work with this?
Any MCP client that supports the streamable HTTP transport: Claude Code, Claude Desktop, Cursor, Windsurf, and other MCP-compatible editors and agents.
What does "fair usage" mean exactly?
A rolling rate limit of 60 requests per minute instead of a hard daily cap. Normal agent workflows never notice it; it only stops runaway loops. Every response includes rate-limit headers so your tooling can see remaining headroom.
Is the lifetime deal really one-time?
Yes. Pay once, keep the key forever, including future updates. The lifetime offer is limited to founding users; after it ends, access moves to a monthly/yearly subscription at the prices shown.
Is the icon data licensed?
Every icon keeps its original open-source license from its source collection, nothing changes there. What you are paying for is hosted API access: uptime, search infrastructure, and support.