ToolRoute
Route each agent task to the few tools it needs (BM25 shortlist). Free tier; x402 or credits.
使うべきか
品質と安全性
ツール定義とプロトコルへの準拠に関する自動分析に基づいています。
コンテキストコスト
これは、サーバーのツールがモデルのコンテキストに読み込まれるたびに消費されるおおよそのトークン数です。数が多いほど、ほかのタスクに使える注意が減ります。
インストール
ワンクリックインストール
これを `claude_desktop_config.json` ファイルに追加してください:
{
"mcpServers": {
"toolroute": {
"url": "https://api.youraitools.co/toolroute/mcp"
}
}
}リモートエンドポイント
https://api.youraitools.co/toolroute/mcpstreamable-httpできること
ツール一覧
ツール（2）
🟡register_tools(tools, url)
Register a tool list once and get a set_id for route_tools. Pass "tools" (array in MCP, OpenAI, Anthropic, Gemini or Bedrock form, max 500) or "url" (public remote MCP server; its tools are listed, never called). Same tools always get the same set_id; sets are kept 90 days after last use. Free.
入力スキーマ
{
"type": "object",
"properties": {
"tools": {
"type": "array",
"items": {
"type": "object"
},
"description": "Tool definitions (MCP, OpenAI, Anthropic, Gemini or Bedrock form), max 500. For route: only if you have no set_id (the set is registered on the fly)."
},
"url": {
"type": "string",
"description": "Register: public remote MCP server URL (https). Tools are listed, never called."
}
}
}🟢route_tools(set_id, tools, query, k, max_k, ...)
Which of my tools does this task need? Send "set_id" (from register_tools) or "tools", plus "query" (the task). Returns a ranked shortlist (adaptive 3-12 tools, or "k"), confidence (high/medium/low/none), matched terms, token savings and the definitions ready to inject ("format": mcp, openai, anthropic, gemini or names). BM25 keyword ranking, no LLM. Free: 200 calls/day per IP, then $0.001 per call via x402 or credits.
入力スキーマ
{
"type": "object",
"properties": {
"set_id": {
"type": "string",
"description": "set_id (\"ts_...\" from register, or a ToolFit pin)."
},
"tools": {
"type": "array",
"items": {
"type": "object"
},
"description": "Tool definitions (MCP, OpenAI, Anthropic, Gemini or Bedrock form), max 500. For route: only if you have no set_id (the set is registered on the fly)."
},
"query": {
"type": "string",
"description": "The task or user request to route, e.g. \"check if this wallet is sanctioned\"."
},
"k": {
"type": "integer",
"minimum": 1,
"maximum": 50,
"description": "Fixed shortlist size (default adaptive)"
},
"max_k": {
"type": "integer",
"minimum": 1,
"maximum": 50,
"description": "Upper bound for the adaptive size (default 12)"
},
"format": {
"type": "string",
"enum": [
"mcp",
"openai",
"anthropic",
"gemini",
"names"
],
"description": "Definition format (default mcp; names = names only)"
}
},
"required": [
"query"
]
}コミュニティ
エビデンス