DA Leads MCP

DA Leads MCP lets Claude, Cursor and other MCP clients query Australian development applications and address-level property intelligence through the DA Leads API.

It exposes 36 tools for DA search, canonical project search and monitoring, nearby applications/projects, council and category lookups, read-only SQL analysis, generic entitlement-filtered property intelligence, the closed public Full Property contract, focused Noise, Flood, Bushfire and Walkability screening, score-free Property Core, Suburb Signals, SA2 scenarios, Neighbourhood Context, Solar Resource, keyless samples and sandbox addresses.

Install

Run without installing globally:

uvx daleads-mcp

Or install with pipx:

pipx install daleads-mcp
daleads-mcp

Configuration

Paid tools use a DA Leads API key. The property sample tools work without a key.

{
  "mcpServers": {
    "da-leads": {
      "command": "uvx",
      "args": ["daleads-mcp"],
      "env": {
        "DALEADS_API_KEY": "dk_live_xxx"
      }
    }
  }
}

DALEADS_API_KEY is the only environment variable read by this package. Treat it as a secret. It is sent only to the fixed official HTTPS API endpoint at https://daleads.com.au/api.

For Claude Code 2.1+, add the stdio server to the current project with:

claude mcp add --scope project da-leads -- uvx daleads-mcp

Keyless sample tools work immediately. For paid tools, launch Claude Code from an environment that already provides DALEADS_API_KEY; do not commit a real key to the project MCP configuration.

Official Skills

The 10 canonical Agent Skills live in skills/. A repository checkout exposes the same files to Codex through .agents/skills/ and to Claude Code through .claude/skills/; both directories are thin relative symlinks, so there is only one maintained instruction source. The Skills guide agents through subject identity, coverage, attribution and product-specific limits. They do not grant an API entitlement or expand data rights.

Tools

Development applications

search_das — Search Australian development applications with filters, newest first. Parameters: state (NSW, VIC, QLD, SA, WA, TAS, NT, ACT), council (full council name, e.g. City of Melbourne), category (trade category, e.g. Renovation / Extension), suburb, postcode, since (ISO date YYYY-MM-DD, lodged on or after), status_group (pending, advertised, approved, rejected, other), is_residential (bool), page (default 1), limit (per page, default 20, max 100). Returns {data: [DA record, ...], meta: {total, page, per_page, pages}}. Each record carries id, address, suburb, postcode, council, state, trade_category, lodgement_date, status, status_group, latitude, longitude, info_url and document links; paid plans add description, summary, sub_category, cost_of_development, decision fields, number_of_dwellings, building_type, storeys and the full documents list. Applicant names are never returned on any plan.

get_da — Retrieve one development application in full. Parameters: da_id (integer record id, as returned by search_das or nearby_das). Returns {data: DA record} with the same field set as above.

nearby_das — Find development applications within a radius of a point, nearest first. Parameters: lat, lng (WGS84 decimal degrees), radius_km (default 5, max 50), category, since, status_group, page, limit (max 100). Returns the same {data, meta} shape as search_das, with an extra distance_km on each record.

list_categories — List every trade category with its record count, for discovering valid category values. No parameters. Returns {data: [{name, slug, is_residential, record_count}, ...], meta: {total}} — categories such as Renovation / Extension, Swimming Pool / Spa, Granny Flat / Secondary Dwelling and Demolition.

list_councils — List councils with coverage and freshness, for discovering valid council values. Parameters: limit (max councils to return, default 50). Returns {data: [{council, state, record_count, last_lodgement, last_fetched}, ...], meta}.

get_stats — Coverage summary for the whole dataset. No parameters. Returns {data: {total_records, records_last_7_days, date_range: {earliest, latest}, by_state: [...], by_category: [...]}}.

sql_query — Run a read-only SQL query for custom aggregation the filter tools cannot express. Pro plan only. Parameters: query (a single SELECT against the das table), params (optional list of values for %s placeholders). Available columns: id, address, address_suburb, address_postcode, council_name, council_reference, state, trade_category, sub_category, application_type, is_residential, lodgement_date, status, cost_of_development, decision_date, decision_status, on_notice_from, on_notice_to, number_of_dwellings, lot_count, land_use, building_type, storeys, latitude, longitude, data_source, date_fetched, documents, info_url. Description and summary are deliberately unavailable here because council free text can contain personal contact details. SELECT only, capped at 1000 rows with a 10 second timeout. Returns {columns: [...], rows: [[...], ...], row_count, truncated}.

Project intelligence

These tools use the same DALEADS_API_KEY as every other paid tool. The API enforces the key's project_intelligence entitlement and rights-cleared field projection; the MCP package does not maintain a second capability list.

search_projects — Search canonical projects rather than independent DA records. Parameters: q, state, normalized stage, normalized project_type, changed_since (ISO-8601 timestamp), page, and limit (max 100). The client sends the provider's canonical per_page parameter. Returns a {data, meta} envelope with contract_version: project-intelligence-v1.

get_project — Retrieve one canonical project. Parameter: project_uid (the stable identifier returned by search_projects). The current v1 response contains identity, normalized status, location, linked applications and project relations. Rights-gated fields are omitted rather than represented as coverage or conflict metadata.

get_project_changes — Retrieve durable changes for one project. Parameters: project_uid, optional initial since, continuation cursor, and limit (max 100). Persist meta.cursor after every response. Use a non-null meta.next_cursor only to fetch the next page immediately. Change values remain redacted until field-level delivery provenance is available.

nearby_projects — Search canonical projects around a WGS84 point. Parameters: lat, lng, radius_km, optional stage and project_type, page, and limit.

create_project_watch — Create or idempotently replay a persistent signed callback for future rights-cleared events on one project_uid. Parameters: project_uid, a public HTTPS callback_url, and a stable idempotency_key. This tool changes external state and returns the signing secret.

list_project_watches — List watches owned by the current API key. Signing secrets are never returned by the list operation.

deactivate_project_watch — Stop one owned watch by watch_uid. Delivery audit rows remain durable. Pending work is suppressed, but an HTTPS request already in flight cannot be recalled and remains visible in the audit. Receivers must deduplicate the stable callback Idempotency-Key header.

The repository includes skills/daleads-project-monitoring for cursor polling and explicit callback management. It is installed from the GitHub repository, not bundled inside the PyPI wheel.

Property intelligence

property_intelligence — Generic entitlement-filtered address-level property lookup for existing and manually issued keys. Parameters: address (free text, e.g. 34 Mary St Clayton VIC) or lat + lng together, and optional components (comma-separated subset to request). A requested name never grants a capability: the API intersects it with the key's server-side entitlement and field-level rights policy. This tool is not proof of a Full Self-Serve entitlement; new Full workflows should use the closed tool below.

full_property_intelligence — Closed public Full Property Intelligence Self-Serve lookup. Parameters: address (recommended) or lat + lng together. It requests the base facts das, poi, planning, transport, utilities, administrative and public_housing; named hazard leaves for bushfire, flood, coastal, landslide and fire history; and named environmental leaves for soil, protected-area, heritage, threatened-species, Ramsar, koala, ecological, EPBC, bioregion and biodiversity context. The exact environmental leaf names are pinned in contracts/focused-api-v1.json.

It also requests exactly these score leaves: scores.noise, scores.aircraft_noise, scores.flood, and scores.bushfire. It never requests top-level hazards, environment or scores, so a new provider field cannot enter Full automatically. Explicit official, rights-gated contamination and register facts may return as Property Core context; they do not grant scores.contamination or publish the standalone Contamination product. Solar, Neighbourhood Heat, Landscape Openness and the Walkability Pilot are excluded. The Full key carries separate property_core and suburb_intelligence capabilities, which do not add fields to this response. Provider-side source and rights gates still apply to every returned field.