The daemon auto-starts on browser commands. gsd-browser daemon health reports state and does not start a session. Use gsd-browser daemon start only when you want to pre-warm or verify daemon lifecycle explicitly.
Always re-snapshot after page changes. Refs are versioned (@v1:e1, @v2:e3). After navigation, form submission, or dynamic content loading, old refs are stale. Run gsd-browser snapshot to get fresh refs before interacting.
Use --json when parsing output. Use text mode when reading output yourself. Use --json when you need to extract values programmatically.
Positional args have no flag prefix. Commands like click, type, hover take positional args. Do NOT add --selector:
gsd-browser click "button.submit" (correct)gsd-browser click --selector "button.submit" (WRONG)
Core workflow pattern: Every browser automation follows: navigate -> snapshot -> interact -> re-snapshot (after DOM changes).
gsd-browser navigate https://example.com
gsd-browser snapshot
# Read snapshot output: @v1:e1 [input type="email"], @v1:e2 [button] "Submit"
gsd-browser fill-ref @v1:e1 "user@example.com"
gsd-browser click-ref @v1:e2
gsd-browser wait-for --condition network_idle
gsd-browser snapshot # REQUIRED - old refs are now stale
Command chaining: Use && when you don't need intermediate output. Run separately when you need to parse output first (e.g., snapshot to discover refs, then interact).
Use the live viewer when the user wants to watch or direct the browser. gsd-browser view opens a localhost viewer with live frames, narrated action history, ref overlays, and pause/step/resume/abort controls. Keep using CLI commands for actions; the viewer is the shared screen and control surface.
Global options available on all commands:
| Flag |
Purpose |
--json |
Structured JSON output |
--browser-path <path> |
Path to Chrome/Chromium |
--cdp-url <url> |
Attach to an already-running Chrome instance |
--session <name> |
Named session for parallel instances |
--no-narration-delay |
Skip narration lead-time sleeps while keeping history/events |
Based on what the user needs, read the appropriate workflow:
| User intent |
Workflow |
| Navigate, click, type, fill forms, interact with pages |
workflows/navigate-and-interact.md |
| Share the browser screen, narrate actions, pause/step/resume/abort |
workflows/live-viewer-and-narration.md |
| Scrape data, extract content, read page structure |
workflows/scrape-and-extract.md |
| Test pages, run assertions, visual regression, mock network |
workflows/test-and-assert.md |
| Debug issues, check logs, diagnose problems |
workflows/debug-and-diagnose.md |
| Install, configure, set up sessions |
workflows/setup-and-configure.md |
After reading the workflow, follow it. Load references only when the workflow directs you to.
All domain knowledge in references/:
Commands: command-reference.md (complete command syntax)
Snapshots: snapshot-and-refs.md (versioned refs, snapshot modes)
Intents: semantic-intents.md (15 predefined intents for find-best/act)
Errors: error-recovery.md (common errors and fixes)
Config: configuration.md (TOML config, env vars, 5-layer merge)
| Workflow |
Purpose |
| navigate-and-interact.md |
Page navigation, clicking, typing, forms, intents |
| live-viewer-and-narration.md |
Live shared viewer, narrated history, refs overlay, controls |
| scrape-and-extract.md |
Data extraction, accessibility tree, page source |
| test-and-assert.md |
Assertions, visual regression, network mocking, test generation |
| debug-and-diagnose.md |
Console/network logs, timeline, debug bundles |
| setup-and-configure.md |
Installation, configuration, sessions, daemon management |
Browser automation task is complete when:
- Target page state is achieved and verified (via assertions or visual confirmation)
- Daemon is stopped if no further browser work is needed (
gsd-browser daemon stop)
- Extracted data is returned in the expected format
- Any saved state (auth, cookies) is persisted for reuse if appropriate
1---2name: gsd-browser-23description: Native Rust browser automation CLI for AI agents. Use when the user needs to interact with websites, navigate pages, fill forms, click buttons, take screenshots, share a live browser view, narrate browser actions, extract structured data, run assertions, test web apps, mock network requests, or automate any browser task. Triggers include "open a website", "fill out a form", "take a screenshot", "show me the browser", "share the screen", "pause the browser", "step through this", "scrape data", "test this web app", "login to a site", "visual regression test", or any task requiring programmatic web interaction.4---56<essential_principles>78**The daemon auto-starts on browser commands.** `gsd-browser daemon health` reports state and does not start a session. Use `gsd-browser daemon start` only when you want to pre-warm or verify daemon lifecycle explicitly.910**Always re-snapshot after page changes.** Refs are versioned (`@v1:e1`, `@v2:e3`). After navigation, form submission, or dynamic content loading, old refs are stale. Run `gsd-browser snapshot` to get fresh refs before interacting.1112**Use `--json` when parsing output.** Use text mode when reading output yourself. Use `--json` when you need to extract values programmatically.1314**Positional args have no flag prefix.** Commands like `click`, `type`, `hover` take positional args. Do NOT add `--selector`:15- `gsd-browser click "button.submit"` (correct)16- `gsd-browser click --selector "button.submit"` (WRONG)1718**Core workflow pattern:** Every browser automation follows: navigate -> snapshot -> interact -> re-snapshot (after DOM changes).1920```bash21gsd-browser navigate https://example.com22gsd-browser snapshot23# Read snapshot output: @v1:e1 [input type="email"], @v1:e2 [button] "Submit"24gsd-browser fill-ref @v1:e1 "user@example.com"25gsd-browser click-ref @v1:e226gsd-browser wait-for --condition network_idle27gsd-browser snapshot # REQUIRED - old refs are now stale28```2930**Command chaining:** Use `&&` when you don't need intermediate output. Run separately when you need to parse output first (e.g., snapshot to discover refs, then interact).3132**Use the live viewer when the user wants to watch or direct the browser.** `gsd-browser view` opens a localhost viewer with live frames, narrated action history, ref overlays, and pause/step/resume/abort controls. Keep using CLI commands for actions; the viewer is the shared screen and control surface.3334**Global options** available on all commands:3536| Flag | Purpose |37|------|---------|38| `--json` | Structured JSON output |39| `--browser-path <path>` | Path to Chrome/Chromium |40| `--cdp-url <url>` | Attach to an already-running Chrome instance |41| `--session <name>` | Named session for parallel instances |42| `--no-narration-delay` | Skip narration lead-time sleeps while keeping history/events |4344</essential_principles>4546<routing>4748Based on what the user needs, read the appropriate workflow:4950| User intent | Workflow |51|-------------|----------|52| Navigate, click, type, fill forms, interact with pages | `workflows/navigate-and-interact.md` |53| Share the browser screen, narrate actions, pause/step/resume/abort | `workflows/live-viewer-and-narration.md` |54| Scrape data, extract content, read page structure | `workflows/scrape-and-extract.md` |55| Test pages, run assertions, visual regression, mock network | `workflows/test-and-assert.md` |56| Debug issues, check logs, diagnose problems | `workflows/debug-and-diagnose.md` |57| Install, configure, set up sessions | `workflows/setup-and-configure.md` |5859**After reading the workflow, follow it. Load references only when the workflow directs you to.**6061</routing>6263<reference_index>6465All domain knowledge in `references/`:6667**Commands:** command-reference.md (complete command syntax)68**Snapshots:** snapshot-and-refs.md (versioned refs, snapshot modes)69**Intents:** semantic-intents.md (15 predefined intents for find-best/act)70**Errors:** error-recovery.md (common errors and fixes)71**Config:** configuration.md (TOML config, env vars, 5-layer merge)7273</reference_index>7475<workflows_index>7677| Workflow | Purpose |78|----------|---------|79| navigate-and-interact.md | Page navigation, clicking, typing, forms, intents |80| live-viewer-and-narration.md | Live shared viewer, narrated history, refs overlay, controls |81| scrape-and-extract.md | Data extraction, accessibility tree, page source |82| test-and-assert.md | Assertions, visual regression, network mocking, test generation |83| debug-and-diagnose.md | Console/network logs, timeline, debug bundles |84| setup-and-configure.md | Installation, configuration, sessions, daemon management |8586</workflows_index>8788<success_criteria>8990Browser automation task is complete when:91- Target page state is achieved and verified (via assertions or visual confirmation)92- Daemon is stopped if no further browser work is needed (`gsd-browser daemon stop`)93- Extracted data is returned in the expected format94- Any saved state (auth, cookies) is persisted for reuse if appropriate9596</success_criteria>
gsd-build