Changelog
A record of product updates, data quality improvements, and new features across the PredictHQ platform. Updated each release cycle.
New FeatureEvents APIFeatures APIBeamForecasts APISaved Locations
Last updated
Was this helpful?
A record of product updates, data quality improvements, and new features across the PredictHQ platform. Updated each release cycle.
The PredictHQ MCP server now exposes tools across the full public API surface, including Events, Broadcasts, Features, Saved Locations, Beam, Forecasts, Predicted Impact Area, and Places & Geocoding.
Previously limited to event search, the MCP server now supports the complete PredictHQ integration workflow through natural language: create Saved Locations, run Beam analyses, retrieve ML-ready features, build and train forecast models, and query impact areas - all without writing API calls directly. Works with any MCP-compatible client including Claude, ChatGPT, Cursor, and Claude Code.
Bolt is now available to all users in the PredictHQ WebApp.
Bolt is an AI-native developer environment that guides you through the right PredictHQ workflows for your use case and produces production-ready integration code. Describe what you're building and Bolt handles the rest — Saved Locations, Beam analysis, calibrated features, forecasts — following PredictHQ best practices throughout.
School holidays for Denmark have been expanded from national-level to municipality-level coverage, reflecting how school holidays are determined locally in Denmark. This applies to all future school holidays and historical data back to 2016, adding 8,900+ events to the dataset. Customers using Danish school holiday data will see increased granularity in event results; historical data has been backfilled to 2016.
Predicted Impact Patterns for the Restaurant industry have been rebuilt for US public holidays, observances, and school holidays using data-driven analysis of real restaurant demand data. This improves forecast accuracy for customers in the restaurant and quick-service retail sectors, particularly around key holiday periods.
The PredictHQ MCP server now supports OAuth session management and token refresh, and is compatible with Claude connectors (previously only ChatGPT was supported). Event search parameter validation has also been improved. The MCP server allows AI agents and LLMs to query PredictHQ event data directly without building custom API integrations.
Descriptions have been added at scale to attended events that previously had none - approximately 79% of the catalogue. Higher-ranked events were prioritised. This improves the usefulness of event data for customers building AI applications, search, and recommendation features where event context matters beyond title and category. More descriptions to come.
Predicted Impact Area is now generally available across Events API, Features API, Beam. It replaces Suggested Radius as the recommended approach for defining the geographic catchment area around a business location. Unlike a simple radius, Predicted Impact Area uses a data-driven model to define the area where events actually influence demand - accounting for real-world geography. The Suggested Radius endpoint remains available but is no longer the recommended default.
Saved Locations now support closed_days and operating_hours fields via the API and WebApp. Beam and the Forecasts API treat closed days as non-demand days, ensuring that forecasts are not distorted by days when a location is not trading. This is particularly relevant for customers with locations that have non-standard trading patterns - seasonal closures, variable hours, or planned shutdowns.
Last updated
Was this helpful?
Was this helpful?