Skip to content

Releases

v1.0.8: corrected run-list field names; ID/failure-mode docs

Choose a tag to compare

@kmatata kmatata released this 23 Aug 17:00

get-actor-run-list descriptions now point at the actual entry fields (id, actId, defaultDatasetId, defaultKeyValueStoreId) instead of a nonexistent runId key, disclose that the list is account-wide, and the README documents where every tool input comes from plus how each fails when wrong.

v1.0.7: non-catalog calls return immediately; lost runs recoverable

Choose a tag to compare

@kmatata kmatata released this 23 Aug 16:42

custom_search / recent_activity orders now return a runId within seconds instead of blocking 30-45s for data that cannot exist yet (the cause of Inspector/Glama timeout errors — the run itself was never affected). New no-ID discovery tools (get-actor-run-list, get-dataset-list, get-key-value-store-list/keys, get-actor-run-log) let any client recover the IDs of a run whose first response was lost. Catalog instant samples are unchanged.

v1.0.6: full five-tool surface under any credentials

Choose a tag to compare

@kmatata kmatata released this 23 Aug 00:52

tools/list now serves the actor tool plus get-actor-run, get-dataset-items, get-key-value-store-record and abort-actor-run from static definitions, so discovery is identical with or without valid credentials.

v1.0.5: bounds renders as real fields

Choose a tag to compare

@kmatata kmatata released this 22 Aug 22:10

The bounding-box input now declares its four number properties with a required set, so strict MCP inspectors render actual input fields instead of an opaque object.

v1.0.4: strict-valid array schemas

Choose a tag to compare

@kmatata kmatata released this 22 Aug 21:59

dedupZpids/dedupMlsIds now declare integer/string item types, satisfying strict JSON Schema validators (MCP Inspector no longer warns).

v1.0.3: instant samples return rows inline

Choose a tag to compare

@kmatata kmatata released this 22 Aug 21:20

Small catalog/recent_activity samples now return the actual listing rows in the same tool response as the run status, so a first try shows leads immediately instead of a two-step fetch. Live-collection orders keep the documented fire/poll/fetch pattern.

v1.0.2: static tool discovery

Choose a tag to compare

@kmatata kmatata released this 22 Aug 02:42

tools/list is now answered locally from a vendored input schema, so tool discovery works even without valid credentials (directory build checks, offline inspection). Actual actor calls are still forwarded to Apify's actors-mcp-server.

v1.0.1: one-tool MCP server

Choose a tag to compare

@kmatata kmatata released this 22 Aug 02:36

One-tool MCP server exposing germane_binoculars/zillow-leads-property-data (agent contacts, price/tax history, foreclosure flags) over stdio. Includes Dockerfile for containerized hosting; --apify-token argument support alongside the APIFY_TOKEN env var.