Now with screen-aware context capture

The context & memory
layer for every model.

Zev captures context from your screen, remembers what matters, and lets you query any LLM — GPT, Claude, Gemini, or open models — with your own API keys.

Windows, macOS & Linux · Bring your own keys · No lock-in

zev · new chat
claude-4-sonnet
3 memories in context
Summarize the Figma flow I was just looking at and compare to the spec from Tuesday's call.
Zev · The checkout flow drops the address step from the spec. Copy tone matches the onboarding memory, but the CTA hierarchy diverges — I flagged three screens…
Features

One layer between you and every model.

Zev is the connective tissue for how you actually work with AI — capture, remember, query, compare.

Screen-aware capture

Zev sees what you see. Highlight a window, a doc, a Figma frame — it becomes context, instantly.

Persistent memory

Every project, meeting, and thread lives in a memory graph. Recall the right context without prompting for it.

Bring your own keys

Plug in OpenAI, Anthropic, Google, or OpenRouter. Your keys, your models, your data.

Any model, one interface

Swap models mid-thread. Compare answers side-by-side. Route the hard stuff to the smart stuff.

Instant test bench

Prompt once, run across models. See latency, cost, and quality — decide with data.

Private by default

Memory stays local unless you say otherwise. No training on your data. Ever.

Models

Your keys.
Every model.

Paste an API key — or connect OpenRouter and unlock hundreds of models in one shot. Switch per-thread, per-message, per-task.

OpenAI
GPT-5, o-series
Anthropic
Claude 4 family
Google
Gemini 3 Pro & Flash
OpenRouter
300+ models, one key
Meta
Llama, hosted or local
Mistral
Large, Codestral
xAI
Grok
DeepSeek
V3, R1
How it works

From your screen to any model in seconds.

01

Install & connect

Download Zev. Paste your API keys or connect OpenRouter for full model access.

02

Capture context

Hit the hotkey. Zev captures the screen, doc, or window you're working in as structured context.

03

Query with memory

Ask anything. Zev pulls the relevant memories and sends the right context to the model you pick.

04

Test & compare

Run the same prompt across models. See what's fastest, cheapest, and best — then choose.