tokenstead.ai

{
  "name": "search_ai_models",
  "kind": "answer",
  "impl": "imperative",
  "description": "Search Tokenstead's catalog of AI models (open-weight and cloud). Filter by free-text query, use case, open weights, or a max input price. Returns name, parameter count, context window, license, scores, and the cheapest verified API price per model.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "q": {
        "type": "string",
        "description": "Free-text name search, e.g. 'qwen', 'glm', 'kimi'"
      },
      "use_case": {
        "type": "string",
        "enum": [
          "coding",
          "reasoning",
          "writing",
          "chat",
          "general",
          "vision"
        ],
        "description": "Primary use case filter"
      },
      "open_weights": {
        "type": "boolean",
        "description": "true = only models with downloadable weights"
      },
      "max_input_price": {
        "type": "number",
        "description": "Only models whose cheapest input API price is at or below this USD per 1M tokens"
      },
      "limit": {
        "type": "number",
        "description": "Max results (default 10, max 25)"
      }
    }
  }
}

{
  "name": "get_model_details",
  "kind": "answer",
  "impl": "imperative",
  "description": "Get full details for one AI model by slug: scores, context window, license, local quantized variants with their memory requirements, and cheapest verified per-token API pricing.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "slug": {
        "type": "string",
        "description": "Model slug, e.g. 'glm-5-2', 'qwen3-8-27b' (from search_ai_models)"
      }
    },
    "required": [
      "slug"
    ]
  }
}

{
  "name": "find_models_for_hardware",
  "kind": "answer",
  "impl": "imperative",
  "description": "Find which AI models can run locally on given hardware, with a fit verdict and estimated tokens per second. Provide a chip name, or an explicit memory size plus memory type. Same engine as Tokenstead's rig finder.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "chip_name": {
        "type": "string",
        "description": "Chip name, e.g. 'RTX 4090', 'M5 Max', 'Ryzen AI Max+ 395'"
      },
      "memory_gb": {
        "type": "number",
        "description": "Memory available in GB (use with memory_type when no chip_name)"
      },
      "memory_type": {
        "type": "string",
        "enum": [
          "unified",
          "vram",
          "system-ram"
        ],
        "description": "Memory type when no chip_name given"
      },
      "use_case": {
        "type": "string",
        "enum": [
          "coding",
          "reasoning",
          "writing",
          "chat",
          "general",
          "vision"
        ],
        "description": "Rank results for this use case"
      },
      "limit": {
        "type": "number",
        "description": "Max results (default 10, max 25)"
      }
    }
  }
}

{
  "name": "estimate_api_cost",
  "kind": "answer",
  "impl": "imperative",
  "description": "Estimate the USD cost of an API workload against one catalog model, at its cheapest verified per-token rate. Napkin math - the /calculator page is the full workload estimator.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "model_slug": {
        "type": "string",
        "description": "Model slug (from search_ai_models)"
      },
      "input_tokens": {
        "type": "number",
        "description": "Total input (prompt) tokens. Default 1,000."
      },
      "output_tokens": {
        "type": "number",
        "description": "Total output (completion) tokens. Default 0."
      }
    },
    "required": [
      "model_slug"
    ]
  }
}

{
  "name": "search_guides",
  "kind": "answer",
  "impl": "imperative",
  "description": "Search Tokenstead's published guides (local-AI hardware, model runbooks, cost breakdowns, benchmark reports).",
  "inputSchema": {
    "type": "object",
    "properties": {
      "q": {
        "type": "string",
        "description": "Free-text title search, e.g. 'dgx spark', 'glm', 'budget'"
      },
      "limit": {
        "type": "number",
        "description": "Max results (default 10, max 25)"
      }
    }
  }
}