BoltAI is the nearest existing UX to OpenCues: its Inline AI lets you type a trigger like gpt: in any macOS text field and the response is inserted in place. OpenCues takes that idea further and makes the typed trigger a full grammar rather than a one-shot prompt.
What BoltAI does well
A capable native Mac AI client at a one-time price, with bring-your-own keys across hundreds of models including local ones, plus customisable inline commands and a hotkey chat window.
Where BoltAI struggles
The inline trigger works by monitoring keystrokes through accessibility permissions, a broad read surface, and the write-back is an injected paste, so results are one-shot insertions rather than managed spans. Rewrite-in-place of existing text is not documented, and as of mid-2026 the inline and command features ship in the v1 app while the v2 rewrite is still absorbing them. macOS-centric, closed source.
By the numbers
BoltAI is the closest on interaction cost: two steps to an answer (type gpt:, hit the trigger hotkey) against OpenCues' one, both with zero steps back to work. The gap shows after the result lands: BoltAI's insertion unwinds with the app's undo, wholesale, where OpenCues keeps each word individually revertable, and BoltAI surfaces nothing unprompted. The counts also measure different ranges: BoltAI's trigger runs one prompt, while the same _ routes free-form fills, in-place transforms, live data from scripts and APIs, and settings, and the capability grid is not close to even. Counts and sequences in the full comparison.
How OpenCues differs
The _ is not a prompt runner: the same trigger does free-form fills, imperative in-place transforms, stateful blanks that pull live data from APIs and scripts, and settings control, with cued word alternatives on plain text. OpenCues is Apache-2.0 open source, reaches terminals, coding agents and Chrome, and can use your existing Claude or ChatGPT plan.
BoltAI reaches any native Mac app, which OpenCues integrations do not: they cover terminals, coding agents and Chrome. OpenCues answers with per-word revertable edits, PII tokenisation, and per-feature model routing.
OpenCues is also extensible by design: cues and blanks are markdown files you write yourself, and blanks can bind shell scripts or runtime classes, all under an open standard anyone can implement. See write your own cue and write your own blank.
Related: Full comparison · Cues vs blanks · BoltAI