Intent Preview
Showing what an agent action will do before it does it, so a human can approve or cancel.
- term
- Intent Preview
- category
- optimization
- short_def
- Showing what an agent action will do before it does it, so a human can approve or cancel.
- long_def
- An AX trust pattern: rather than acting silently, the agent surfaces the planned call and its effect ('this will POST and write one record') for confirmation. Pairs with an action audit log of what actually happened.
- see_also
action-auditagent-experience- etymology_origin
- — verify-against-primary-at-build ↗ https://biilmann.blog/articles/introducing-ax/ — 'intent preview' is an Agent Experience (AX) trust pattern; confirm whether Biilmann/Netlify or this Almanac is the first named source for the exact term
- related_to
action-auditagent-experience- contrast_with
- Unlike an action audit log, which records what already happened (the 'what did' half), intent preview shows what is about to happen (the 'what will' half) so a human can approve or cancel first.
- example
- Before an agent writes data, an intent preview surfaces the planned call and its effect — 'this will POST and write one record' — for human confirmation.
- source
- https://biilmann.blog/articles/introducing-ax/
- status
- emerging
- why_it_matters
- Intent preview is a core AX trust pattern: it keeps a human in the loop for consequential agent actions without forcing the agent to stop for trivial ones.
- sameAs
- —
- bridge_entity
- agent-readiness
- last_verified
- 2026-06-15
- md_twin
- /glossary/intent-preview.md