koko-fetch is a CLI built on the Koko SDK. It connects to a running CDP server and fetches page HTML or structured extracts — useful for quick probes, CI checks, and shell pipelines.
Prerequisites
Koko must be reachable over CDP. Either launch from the SDK first, or start a server:
# Option A: SDK launch (see Quick start)
# Option B: long-lived server
koko serve --host 127.0.0.1 --port 9222Fetch HTML
KOKO_CDP=http://127.0.0.1:9222 npx koko-fetch https://example.comDefault wait: domcontentloaded. HTML is written to stdout.
Structured extract (JSON)
KOKO_CDP=http://127.0.0.1:9222 npx koko-fetch \
https://en.wikipedia.org/wiki/Earth --extractThe extract path uses the SDK page.extract() helper — optimized for crawler workloads with TTFX probe and structured fields (title, links, etc.).
Environment variables
| Variable | Description |
|---|---|
KOKO_CDP | CDP HTTP endpoint (e.g. http://127.0.0.1:9222) |
When to use SDK instead
- Multi-step flows (login, form fill, click chains)
- Agent APIs (
semanticTree,NodeHandle) - Session state capture/restore
- Profile launch via
Browser.launch()