rep-cli

rep is the command-line companion to the rep+ browser extension. It lets coding agents such as Claude Code and Codex drive your real Arc or Chrome profile, capture its network traffic with response bodies, and inspect or replay requests. The CLI reaches the extension through a native-messaging host (rep-host), and captures stay on your machine.

• Capture: open pages or run page actions; archive HTTP bodies, WebSocket messages and opt-in WebTransport/WebRTC data-channel payloads per task. See transport capture.
• Packets and media: collect filtered packets through native libpcap, or launch private Chromium diagnostics for TLS/QUIC keys and original WebRTC RTP. Track recordings remain available separately. See native capture.
• Inspect: summary, context, list, search, body, and stream return bounded views with capture status.
• Act: replay or mutate captured requests, or run verified UI steps with browser interact.
• Isolate: each agent works in its own workspace/task, so captured data never mixes.
• Record: opt into an explicit run journal with --run; retain failed/unknown operations, saved artifacts and native diagnostic files.

Install

Requires macOS, Go 1.25+, and Arc or another Chromium browser.

git clone https://github.com/sijan2/rep-cli && cd rep-cli
scripts/build_install.sh --host    # runs tests, installs rep and rep-host to ~/.local/bin

Make sure ~/.local/bin is on your PATH. Then add the extension and connect it:

git clone https://github.com/sijan2/rep ~/rep
# arc://extensions or chrome://extensions → Developer mode → Load unpacked → ~/rep
rep browser install --extension-path ~/rep                    # Arc (default)
rep browser install --extension-path ~/rep --browser chrome   # Chrome
# Reload rep+, then check the bridge:
rep browser doctor

Use

Data commands need a workspace and a task; give each agent its own task. Browser commands use Arc unless you pass --browser chrome.

export REP_WORKSPACE=demo REP_TASK=first-look

rep browser open https://example.com --keep-tab   # navigate in a background tab and capture
rep summary --max-bytes 4096                      # grouped overview with a cursor
rep context --since CURSOR                        # only what changed since that cursor
rep primary example.com                           # mark the target domain
rep list --api                                    # API calls to primary domains
rep body ID --info                                # check completeness, then --head or --pointer
rep replay ID --set 'header.X-Debug=1'            # re-send with a change and diff

rep agent-prompt prints paste-ready instructions for an agent, and rep describe <command> prints any command's full contract.

Optional:

• rep browser headless start --extension ~/rep runs a private headless profile per task (needs Chrome for Testing or Chromium). Add --browser headless to later commands.
• rep jev config --env-file /path/to/.env enables Jev element lookup (browser select) and traffic classification. Needs a TypeSafe API key.

For durable workflow evidence:

rep evidence begin --intent 'Inspect local fixture' --stop 'Save observations'
rep --run RUN_ID browser screenshot --tab TAB_ID --raw-json
rep evidence operations RUN_ID --limit 10
rep evidence import RUN_ID trace.bin --manifest collection.json

Runs link explicit browser operations and artifacts. Native diagnostic imports preserve bytes and declared collector metadata; they do not start live platform tracing. Network capture remains serialized against page/semantic execution on the same tab. See run evidence for the supported commands and architecture for protocol coverage, limits and measured local lookup/durability costs.

Docs

Workspaces and context · Body capture · Interactions · Headless · Jev · Decision runtime · Run evidence · Architecture · Arc control · Android

Develop

go test ./...