For AI agents and copilots
ToolHund helps AI agents choose the right AI service for any task.
One decision endpoint. One MCP tool. One licensed data feed. Built for apps, agents, copilots and workflows that need to pick an AI tool at runtime — not read a listicle.
Recommend Engine · v1 · Developer $49/mo
AI agents need tools too.
One endpoint that turns a natural-language task into a ranked, reasoned pick — with pricing, tradeoffs and monetization disclosure. Ship it inside your app, agent or copilot in an afternoon.
Example queries
"best AI voice API for Turkish narration under $50/mo""best AI video tool for lip-sync ads""best AI search API for a research agent"
Stripe checkout. API key provisioned manually within 24 hours after payment — you'll get an email once it's live.
Request data licenseView OpenAPIConnect via MCPWhat agents get
Recommend API
POST /api/v1/recommend — natural-language task → 1–3 ranked AI tools with reasoning, pricing and tradeoffs. Deterministic, no LLM in the ranking path.
See endpointMCP tool: recommend_ai_tool
Drop the Recommend Engine straight into Claude Desktop, Cursor or any MCP-compatible agent. One tool call, ranked picks with monetization disclosure.
Connect via MCPOpenAPI spec
Machine-readable OpenAPI 3.1 for the whole API — recommend, categories, tools, compare, deals, news. Wire it into any codegen or agent framework.
View OpenAPIllms.txt
A curated llms.txt so LLM crawlers discover the endpoints and MCP entrypoints without scraping HTML.
See llms.txtStructured AI tool data
Pricing, categories, alternatives, scores and freshness metadata — normalized fields, not scraped HTML. Available via API or a licensed data feed.
License the dataAttribution & disclosure
Every recommendation carries a monetization block: has_affiliate_link, is_sponsored (always false in ranking), and a plain-English disclosure. Ship it in your UI.
Read methodologyUse cases
Voice agents
Recommend TTS/STT/voice APIs by language, latency and cost inside a call agent.
Video agents
Pick the right lip-sync, avatar or B-roll tool for an incoming ad brief.
Research agents
Route research tasks to the right search / citation API instead of hardcoding one vendor.
SaaS copilots
Let your product suggest the right AI tool for a job — with reasons, not just names.
Newsletters
Power a weekly 'best tool for X' section from a licensed feed instead of manual curation.
Marketplaces
Enrich AI tool listings with normalized categories, pricing and scores.
Copy-paste examples
Everything an agent needs to call ToolHund in under a minute.
1. curl — POST /api/v1/recommend
Public demo, 20 req/day per IP. Add Authorization: Bearer th_live_... for keyed access.
curl -X POST https://toolhund.com/api/v1/recommend \
-H "Content-Type: application/json" \
-d '{
"task": "generate a cinematic short video with lip sync",
"constraints": { "budget_max_usd_month": 50 },
"limit": 3
}'2. Claude Desktop — MCP config
Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or the equivalent on your platform, then restart Claude.
{
"mcpServers": {
"toolhund": {
"url": "https://toolhund.com/mcp"
}
}
}Exposes recommend_ai_tool, search_tools, compare_tools, get_pricing and more.
3. Discovery surfaces
- /openapi.json — OpenAPI 3.1 with full
/api/v1/recommendschema and examples. - /llms.txt — curated site index for LLM crawlers.
- /api/v1/map — machine-readable endpoint catalog including MCP tools.
Connect ToolHund to your agent
ToolHund ships as a public MCP server and a GPT Action. Add it to any of the ecosystems below and your agent gets ranked AI tool picks with reasoning, pricing signals and monetization disclosure.
Central metadata source for MCP servers.
MCP server marketplace for Claude, Cursor and Windsurf.
Curated MCP directory with client compatibility matrix.
Community MCP registry with weekly picks.
Searchable MCP index used by many agent builders.
Index of AI-ready websites publishing llms.txt.
Ship ToolHund as a ChatGPT app via GPT Actions.
Discover / build / use network for AI agents.
GPT Action manifest
Point ChatGPT's custom GPT builder at the manifest and OpenAPI spec — the recommend_ai_tool action is public and requires no auth.
- /.well-known/ai-plugin.json — plugin / action manifest.
- /openapi.json — OpenAPI 3.1 spec.
Copy-paste submission description
ToolHund MCP Server helps AI agents choose the right AI service for a task. It exposes a recommend_ai_tool tool that accepts a natural language task and returns ranked AI tool recommendations with reasoning, pricing signals, tradeoffs, ToolHund Score and relevant links. Example: "best AI voice API for Turkish narration under $50/month" Use cases: - AI agents choosing voice, video, search, coding, automation or memory tools - SaaS copilots recommending third-party AI services - Developers building AI tool discovery into their apps - Marketplaces and newsletters using structured AI service data Endpoint: https://toolhund.com/mcp Docs: https://toolhund.com/for-ai-agents OpenAPI: https://toolhund.com/openapi.json llms.txt: https://toolhund.com/llms.txt Contact: api@toolhund.com
Request a data license
For high-volume use, private data feeds or custom licensing. We'll review and reply by email.