Extcord

Find out what this server can do. With no arguments: every op by category with a one-line intent, plus saved recipes. With query: the best-matching ops and recipes. With op: that op's full contract and examples.

{
  "type": "object",
  "properties": {
    "query": {
      "anyOf": [
        {
          "type": "string"
        },
        {
          "type": "null"
        }
      ],
      "default": null,
      "description": "Words describing what you want to do, e.g. 'html table to rows' or 'days between dates'.",
      "title": "Query"
    },
    "op": {
      "anyOf": [
        {
          "type": "string"
        },
        {
          "type": "null"
        }
      ],
      "default": null,
      "description": "An op id to get its full contract: args schema, behavior, worked examples.",
      "title": "Op"
    },
    "accepts": {
      "anyOf": [
        {
          "type": "string"
        },
        {
          "type": "null"
        }
      ],
      "default": null,
      "description": "Only ops that take this input type: text, html, json, list, table, number, boolean.",
      "title": "Accepts"
    }
  },
  "title": "discoverArguments"
}

{
  "type": "object",
  "additionalProperties": true,
  "title": "discoverDictOutput"
}

Run one operation and get its result as a handle: type, size, estimated tokens, the value itself if small (else a preview), and suggested next ops. Example: invoke(op="fetch", args={"url": "https://example.com"}), then invoke(op="html_to_text", input="<handle from that result>").

{
  "type": "object",
  "properties": {
    "op": {
      "description": "Op id, e.g. 'fetch', 'html_to_text', 'calc'. See discover().",
      "title": "Op",
      "type": "string"
    },
    "input": {
      "default": null,
      "description": "The data to operate on: a handle id from an earlier result (h_...), or a literal value (text, number, list, object). Omit for ops that take no input, like fetch, calc or now.",
      "title": "Input"
    },
    "args": {
      "anyOf": [
        {
          "additionalProperties": true,
          "type": "object"
        },
        {
          "type": "null"
        }
      ],
      "default": null,
      "description": "The op's arguments, e.g. {\"url\": \"https://example.com\"} for fetch or {\"expr\": \"2 ** 10\"} for calc.",
      "title": "Args"
    },
    "input_type": {
      "anyOf": [
        {
          "type": "string"
        },
        {
          "type": "null"
        }
      ],
      "default": null,
      "description": "Override the inferred type of a literal input, e.g. 'html'.",
      "title": "Input Type"
    },
    "inline_max_tokens": {
      "default": 300,
      "description": "Results up to this many estimated tokens include their value directly; larger ones return a preview. Default 300.",
      "maximum": 20000,
      "minimum": 0,
      "title": "Inline Max Tokens",
      "type": "integer"
    }
  },
  "required": [
    "op"
  ],
  "title": "invokeArguments"
}

{
  "type": "object",
  "additionalProperties": true,
  "title": "invokeDictOutput"
}

Chain operations in one call; only the final result comes back, intermediate data stays on the server. Example steps: [{"op": "fetch", "args": {"url": "https://example.com/pricing"}, "as": "page"}, {"op": "extract_tables", "input": "$page", "args": {"index": 0}}, {"op": "to_markdown", "input": "$1"}]. Use dry_run=true first when unsure.

{
  "type": "object",
  "properties": {
    "steps": {
      "description": "Ordered steps: {\"op\": id, \"input\"?: ref|handle|value, \"args\"?: {...}, \"as\"?: name}. Refer to earlier steps as \"$0\", \"$1\", \"$name\", or paths like \"$1[0]\"; refs work inside args too.",
      "items": {
        "additionalProperties": true,
        "type": "object"
      },
      "title": "Steps",
      "type": "array"
    },
    "result_from": {
      "anyOf": [
        {
          "type": "string"
        },
        {
          "type": "null"
        }
      ],
      "default": null,
      "description": "Which step to return, as a ref like \"$2\" or \"$prices[0]\". Default: the last step.",
      "title": "Result From"
    },
    "dry_run": {
      "default": false,
      "description": "Check refs, args, types, cost and side effects without running anything.",
      "title": "Dry Run",
      "type": "boolean"
    },
    "inline_max_tokens": {
      "default": 300,
      "description": "Results up to this many estimated tokens include their value directly; larger ones return a preview. Default 300.",
      "maximum": 20000,
      "minimum": 0,
      "title": "Inline Max Tokens",
      "type": "integer"
    }
  },
  "required": [
    "steps"
  ],
  "title": "run_pipelineArguments"
}

{
  "type": "object",
  "additionalProperties": true,
  "title": "run_pipelineDictOutput"
}

Look at a handle: its summary, its value one page at a time, or its provenance. Check est_tokens in the summary before reading a big value; it's usually cheaper to narrow it with an op first (find, select, filter, fit).

{
  "type": "object",
  "properties": {
    "handle": {
      "description": "A handle id like h_3f9a1c2b7e.",
      "title": "Handle",
      "type": "string"
    },
    "view": {
      "default": "summary",
      "description": "summary: size, preview, next ops. value: the data itself, paged. provenance: the chain of ops, args and sources that produced it.",
      "enum": [
        "summary",
        "value",
        "provenance"
      ],
      "title": "View",
      "type": "string"
    },
    "offset": {
      "default": 0,
      "description": "For view=value: start position (chars for text, items for lists).",
      "minimum": 0,
      "title": "Offset",
      "type": "integer"
    },
    "limit": {
      "anyOf": [
        {
          "maximum": 100000,
          "minimum": 1,
          "type": "integer"
        },
        {
          "type": "null"
        }
      ],
      "default": null,
      "description": "For view=value: page size (default 4000 chars or 50 items).",
      "title": "Limit"
    }
  },
  "required": [
    "handle"
  ],
  "title": "inspectArguments"
}

{
  "type": "object",
  "additionalProperties": true,
  "title": "inspectDictOutput"
}

Run a saved pipeline in one call. Built-ins: page_to_markdown (url, max_tokens), page_outline (url), page_links (url, contains), page_tables (url, index), find_on_page (url, pattern).

{
  "type": "object",
  "properties": {
    "name": {
      "description": "Recipe name, e.g. 'page_to_markdown'. discover() lists them.",
      "title": "Name",
      "type": "string"
    },
    "params": {
      "anyOf": [
        {
          "additionalProperties": true,
          "type": "object"
        },
        {
          "type": "null"
        }
      ],
      "default": null,
      "description": "Values for the recipe's params, e.g. {\"url\": \"https://example.com\"}.",
      "title": "Params"
    },
    "dry_run": {
      "default": false,
      "title": "Dry Run",
      "type": "boolean"
    },
    "inline_max_tokens": {
      "default": 300,
      "description": "Results up to this many estimated tokens include their value directly; larger ones return a preview. Default 300.",
      "maximum": 20000,
      "minimum": 0,
      "title": "Inline Max Tokens",
      "type": "integer"
    }
  },
  "required": [
    "name"
  ],
  "title": "run_recipeArguments"
}

{
  "type": "object",
  "additionalProperties": true,
  "title": "run_recipeDictOutput"
}

Save a working pipeline under a name so any agent can run it later with run_recipe. Saved recipes are shared and can't be overwritten.

{
  "type": "object",
  "properties": {
    "name": {
      "description": "3-64 lowercase letters, digits or underscores, e.g. acme_pro_price.",
      "title": "Name",
      "type": "string"
    },
    "description": {
      "description": "One sentence: what it returns, for other agents searching later.",
      "title": "Description",
      "type": "string"
    },
    "steps": {
      "description": "Pipeline steps, with values to vary written as {{param}}, e.g. {\"op\": \"fetch\", \"args\": {\"url\": \"{{url}}\"}}.",
      "items": {
        "additionalProperties": true,
        "type": "object"
      },
      "title": "Steps",
      "type": "array"
    },
    "params": {
      "anyOf": [
        {
          "additionalProperties": {
            "additionalProperties": true,
            "type": "object"
          },
          "type": "object"
        },
        {
          "type": "null"
        }
      ],
      "default": null,
      "description": "Each {{param}}: {\"type\": \"string\", \"description\": \"...\", \"default\"?: ...}.",
      "title": "Params"
    },
    "result_from": {
      "anyOf": [
        {
          "type": "string"
        },
        {
          "type": "null"
        }
      ],
      "default": null,
      "description": "Ref of the step to return; default the last.",
      "title": "Result From"
    }
  },
  "required": [
    "name",
    "description",
    "steps"
  ],
  "title": "save_recipeArguments"
}

{
  "type": "object",
  "additionalProperties": true,
  "title": "save_recipeDictOutput"
}