jev-rules

Checked, not fully reviewed. We ran its 131 tests — all pass — and read its hooks. No Jev calls.

Standing instructions pile up: test the payment code, use British spelling, follow the deploy checklist. jev-rules asks, per prompt, which of them apply, and injects only those, with a rules pane showing what was picked. On an edit it asks again about the file being edited.

What leaves the machine is worth knowing before you install it. A prompt hook sends { request: prompt }, truncated at 24,000 characters, to api.typesafe.ai/v1/systemone — or to Vercel AI Gateway if the key is a gateway key. The edit hook sends only the file’s path, not its contents, and answers are cached per file. The URL is a constant, so you cannot point it at a local model. It fails open: if the call times out (8 s) or errors, every rule it could not judge is included and the prompt proceeds, which is the right default for a tool that only adds context.

MIT, 64 stars. A small, careful plugin; install it knowing that your prompts take one extra trip.

See how it compares with other tools in Best Jev tools, tested hands-on.

Review updated Oct 4, 2026. Numbers quoted from the project are its author's own; we don't publish our own measurements of Jev.