Voice-driven editing
Local Whisper, sub-100ms turn taking.
Features
eri ships Cursor-grade tab completion, inline edit, composer, agent mode, and codebase chat. Then it adds voice, gaze, diffusion preview, and iteration history — the parts a keyboard alone can't reach.
The full catalog
Cursor-parity across the keyboard, AI surfaces, codebase intelligence, and collaboration — with voice, gaze, diffusion, and iteration history layered on top.
Only in eri
01 / 05
Local Whisper, sub-100ms turn taking.
Patches arrive as a wash, not a crawl.
Every turn is a reversible checkpoint.
Look at the element. eri resolves it.
The keyboard surface
02 / 05
Predictive multi-line, accept with one key.
Describe the change, watch the diff land.
One click to merge a streaming patch.
Suggest commands, run them with consent.
Where the model meets you
03 / 05
Threaded conversation pinned to your repo.
Multi-file edits streamed into one review.
Plan, edit, run, verify — all reviewable.
Drop a screenshot, get a working component.
Persistent scratchpad with live context.
Per-task model selection from one TOML.
Zero retention, no payload leaves your box.
Hand off long jobs to a background runner.
Knowing the repo
04 / 05
Pull files, folders, web, docs into context.
Local embeddings, semantic-grade search.
Live results stitched into the prompt.
Pin @docs to any framework or package.
Ship project guidelines via .erirules.
Bring any tool, keep your own keys.
Continuous review, ranked by blast radius.
Diagnostics rolled into every suggestion.
Out of the editor
05 / 05
Auto commit messages, branch automation.
Open a PR with a real change summary.
Spawn, switch, and clean up by intent.
Read your diffs back with annotated risk.
Next
Features are easier to feel than to read. The demo walks the full surface in under three minutes.