TXLookup developer reference · MCP tools, agents, install, telemetry.

discover_datasets

Find a relevant Texas civic dataset for a question. Returns up to 5 candidates from the curated catalog.

{
  "query": "food truck permits 78702",
  "city": "Austin"
}

get_dataset_schema

Inspect a dataset's columns, types, and field-name aliases before constructing a SoQL query. Always run this BEFORE fetch_data on a dataset you haven't queried recently.

{
  "dataset_id": "3syk-w9eu"
}

fetch_data

Fetch up to 100 rows from a Texas Socrata dataset. SoQL `where` clause required for any scoped query (zip, date range). Cache-first: cached rows return in <50ms; live fetch falls back automatically when cache is cold.

{
  "dataset_id": "3syk-w9eu",
  "where": "original_zip='78702' AND issue_date >= '2026-01-01'",
  "limit": 100
}

ask_data

End-to-end agent loop. Plain-English question → reason → plan → tool dispatch (with replan on failure + doom-loop guard + critic-driven revision) → cited synthesized answer. Wraps the multi-agent orchestrator (data analyst + reporter + critic).

{
  "query": "Permits + violations spike together by zip this year"
}

get_task_status

Poll status of a long-running ask_data task started in async mode.

{
  "task_id": "task_01HZ…"
}

create_miro_board

Create a Miro board for visualizing agent results. Returns board ID + URL.

{
  "name": "78702 permits Q1",
  "description": "Auto-built by ask_data"
}

add_to_miro

Post a structured payload (frame + sticky grid + chart card) to an existing Miro board.

{
  "board_id": "uXjVM…",
  "items": [
    "frame",
    "sticky_grid",
    "chart_card"
  ]
}

list_known_tools

Self-introspection: returns this tool catalog. Useful for agents that load the server dynamically.

{}