Here's a quick look at the latest app updates and fixes.

We've updated the Anthropic (Claude) app to include support for Anthropic's newest and most powerful model, Claude Opus 5. Use it in scenarios that need to reason over large inputs at once, like full codebases, detailed specs, compliance manuals, diagrams, or spreadsheets. Extended thinking is on by default, with low, medium, or high effort settings, and it can act as a manager agent, delegating to Make AI Sub-Agents and running writer/verifier loops. Available on all plans, via your own Anthropic API key or Make AI Provider. For more information, see Anthropic (Claude) documentation.

We've updated the Google Gemini AI app to include support for Google's latest model, Gemini 3.6 Flash. Use it to analyze large text and mixed-media inputs, text, images, audio, or video, and return structured JSON for the next step in your . Supports up to one million input tokens, making it well suited to long transcripts, document sets, or support histories. Available with a Gemini API key in Make AI Toolkit, Google Gemini AI, and Make AI Agent (New). For more information, see Google Gemini AI documentation.

New app version now built on Harvest API v3, using OAuth instead of an API key. Greenhouse Recruiting is shutting down the Harvest v1/v2 API on August 31, 2026. Several modules have updated inputs and outputs to match the new data structure. This replaces the deprecated Harvest v1/v2 API and its API key authentication with an OAuth connection.

Scenarios still using the old app version and API key connection will stop working after that date. Migrate your scenarios to the new app version and an OAuth connection before then to avoid disruption.

See the full module-by-module migration guide, including field mapping and breaking changes, in the Greenhouse Recruiting page.

We've updated the OpenAI (GPT) app to reflect recent pricing reductions for its latest models. This update makes using these powerful models in your scenarios more cost-effective.

The pricing for the gpt-5.6-sol model remains unchanged. No action is needed from you - scenarios using the updated models will automatically consume fewer credits.

To learn more about credit usage in the Simple text prompt module, see OpenAI modules.

An issue that caused the Watch Attendees in Organization trigger to time out for organizations with a large volume of attendees has been resolved. The module is now more reliable for scenarios processing a high number of attendees.

Search Records: An issue where searching for timesheets using a single contract type resulted in an error has been resolved. The module now correctly processes single-item filter criteria

An issue that prevented users from creating and verifying connections has been resolved. You can now connect to Spotify and run your scenarios as expected.

An issue that prevented users from creating or reauthorizing Salesforce Pardot connections when specific security policies (PKCE) were enabled by Salesforce has been resolved. All users can now create and manage their connections without interruption.

An issue that prevented users with newer Atera API tokens from creating a successful connection has been resolved. The app now uses the required authentication method, ensuring all valid tokens work as expected.

Error Handling: When an Atera module failed due to an invalid or expired API key, a generic error was previously displayed. This has been fixed to show a clear "[401] Invalid API Key." message, helping you identify and resolve authentication issues faster.