Hex
Hex is making its AI agent programmable and its generated apps directly editable.
◆Recent moves
- 6d ago
More control over Generative app iteration
⚡ SPARKBYOK (bring your own AI provider) is the headline here — teams can now route Hex Agent calls through their own model credentials, ending Hex's control over which models power your data apps. Combined with direct text editing and targeted chart updates, this release shifts the agent from a black box into something users own end-to-end.
- 20d ago
Chart controls and easier collaboration
Generative chart controls give users finer-grained adjustment over what the agent produces, reducing the round-trip of 'regenerate this chart' prompts. Shareable OAuth results extend collaboration without requiring teammates to re-authenticate. Both moves build on the trajectory toward more user control over AI outputs.
- 1mo ago
Start building an app right from your conversation
The chat-to-notebook toggle flattens the workflow — analysts can now start in conversation mode and graduate to a full notebook without switching contexts. The Hex Agent reading entire Slack threads rather than just the last message meaningfully improves the quality of generated outputs for collaborative discussions.
- 1mo ago
Tune how the agent runs
Fast mode and model visibility (seeing which model Auto picked) address two real friction points: speed and trust. Adding GPT-5.6 variants expands the model roster that the Aug-30 BYOK move will eventually let users override entirely — the model picker is both useful now and a stepping stone to the forthcoming provider abstraction.
- 1mo ago
Command (line) and conquer
⚡ SPARKCLI and API access to the Hex Agent is an architectural move — it opens Hex's AI capabilities to programmatic workflows, CI pipelines, and external integrations that don't go through the browser. Adding Claude Opus 5 alongside this puts a capable model at the end of the API, making the combination immediately useful for automated data app generation.
- 1mo ago
Get hands-on with your generative apps
Letting users edit generated app code directly inside Hex closes the feedback loop between AI output and human intent — the agent no longer has to regenerate when a user wants a small tweak. Desktop notifications and the agent's ability to see its own built apps are operational improvements that reduce context-switching for longer-running agent tasks.