Monica, Sider and Merlin put a chat sidebar next to every web page, bundling many models behind their own credit plans. OpenCues rejects the sidebar as a surface: the text field you are writing in is the interface, and the model you use is your own.
What the sidebars do well
One subscription, many frontier models, page-aware chat, summaries and quick actions, at real scale (millions of installs across the tier).
Where the sidebars struggle
The sidebar is the cost: every interaction is a context switch into a panel and a transfer back out, and the reinsertion step is mostly undocumented copy-paste. The extensions read page content to be page-aware, a broad privacy surface, and inference is metered through each vendor's credit system with no local option.
By the numbers
The sidebar pattern measures six steps to get an answer into your text, three of them the copy, refocus, paste journey back, against one step and zero end-lag inline. The sidebar plus its floating page icon are two standing surfaces; OpenCues adds none. Counts and sequences in the full comparison.
How OpenCues differs
No panel, no credit resale. OpenCues works inline in the field with a typed _, and bills nothing: you bring API keys, your existing Claude or ChatGPT plan, or a local model. It is Apache-2.0 open source and reaches surfaces a browser extension cannot: terminals, coding agents and the shell.
Context differs too: sidebars are page-aware, OpenCues is field-aware, it reads the input you are writing in (and, opt-in, the field's ambient metadata) rather than the whole page.
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 · AI in any text field · Monica