Automate filling web forms including login, registration, checkout, and multi-step form wizards using Python code execution.
\nAll code runs via openbrowser-ai -c. The daemon starts automatically and persists variables across calls. All browser functions are async -- use await.
The CLI daemon also persists cookies and login state in ~/.config/openbrowser/profiles/daemon/storage_state.json, so authenticated sessions can be reused across later runs.
Before running, verify openbrowser-ai is installed:
\nopenbrowser-ai --help\nIf not found, install:
\n# macOS/Linux\ncurl -fsSL https://raw.githubusercontent.com/billy-enrizky/openbrowser-ai/main/install.sh | sh\n\n# Windows (PowerShell)\nirm https://raw.githubusercontent.com/billy-enrizky/openbrowser-ai/main/install.ps1 | iex\nopenbrowser-ai -c - <<'EOF'\nawait navigate(\"https://example.com/login\")\nstate = await browser.get_browser_state_summary()\nprint(f\"Page: {state.title} ({state.url})\")\nprint(f\"Interactive elements: {len(state.dom_state.selector_map)}\")\nEOF\nopenbrowser-ai -c - <<'EOF'\n# List all interactive elements with their indices\nstate = await browser.get_browser_state_summary()\nfor index, element in state.dom_state.selector_map.items():\n tag = element.tag_name\n text = element.get_all_children_text(max_depth=2)[:60]\n placeholder = element.attributes.get(\"placeholder\", \"\")\n input_type = element.attributes.get(\"type\", \"\")\n name = element.attributes.get(\"name\", \"\")\n print(f\"[{index}] <{tag}> type={input_type} name={name} placeholder=\\\"{placeholder}\\\" text=\\\"{text}\\\"\")\nEOF\nopenbrowser-ai -c - <<'EOF'\n# Fill fields using their indices from Step 2\nawait input_text(index=5, text=\"user@example.com\")\nawait input_text(index=7, text=\"secure-password\")\nEOF\nFor fields that need clearing first:
\nopenbrowser-ai -c - <<'EOF'\nawait click(index=5)\nawait evaluate(\"document.activeElement.select()\")\nawait input_text(index=5, text=\"new-value\")\nEOF\nStandard HTML select elements:
\nopenbrowser-ai -c - <<'EOF'\nawait select_dropdown(index=12, text=\"United States\")\nEOF\nTo see available options first:
\nopenbrowser-ai -c - <<'EOF'\noptions = await dropdown_options(index=12)\nprint(options)\nEOF\nCustom dropdown components:
\nopenbrowser-ai -c - <<'EOF'\nawait evaluate(\"\"\"\n(function(){\n const select = document.querySelector(\"select#country\");\n select.value = \"US\";\n select.dispatchEvent(new Event(\"change\", { bubbles: true }));\n})()\n\"\"\")\nEOF\nopenbrowser-ai -c - <<'EOF'\nawait click(index=15) # Click checkbox/radio\n\n# Verify state\nchecked = await evaluate(\"\"\"document.querySelector(\"input[name=agree]\").checked\"\"\")\nprint(f\"Checkbox checked: {checked}\")\nEOF\nopenbrowser-ai -c - <<'EOF'\nawait click(index=20) # Click submit button\nawait wait(2)\n\n# Verify submission\nstate = await browser.get_browser_state_summary()\nprint(f\"After submit: {state.url}\")\nEOF\nOr submit via JavaScript:
\nopenbrowser-ai -c - <<'EOF'\nawait evaluate(\"document.querySelector(\\\"form\\\").submit()\")\nEOF\nopenbrowser-ai -c - <<'EOF'\n# Check for success/error messages\nresult = await evaluate(\"\"\"\n(function(){\n const success = document.querySelector(\".success, .alert-success, [role=\\\"alert\\\"]\");\n const error = document.querySelector(\".error, .alert-danger, .validation-error\");\n return {\n success: success?.textContent?.trim(),\n error: error?.textContent?.trim(),\n url: window.location.href\n };\n})()\n\"\"\")\nprint(result)\nEOF\nopenbrowser-ai -c - <<'EOF'\nfor step in range(1, 5):\n # Discover fields for current step\n state = await browser.get_browser_state_summary()\n print(f\"Step {step}: {len(state.dom_state.selector_map)} elements\")\n\n # Fill fields (indices vary per step)\n # ... fill fields here ...\n\n # Click Next/Continue\n # Find the next button\n for idx, el in state.dom_state.selector_map.items():\n text = el.get_all_children_text(max_depth=1).lower()\n if \"next\" in text or \"continue\" in text:\n await click(index=idx)\n await wait(2)\n break\nEOF\n-c - <<'EOF'), so all Python syntax works without shell escaping issues.browser.get_browser_state_summary() before typing -- do not guess element indices.evaluate() to bypass custom components that do not respond to standard click/type.-c calls while the daemon is running, so you can store field indices in one call and use them in the next.This step is mandatory. Run it after the form submission finishes, whether the submit succeeded or the form rejected the input. Without it, the daemon keeps Chrome running until its 10-minute idle timeout, leaving a stale browser process, a locked profile, and (on macOS/Linux desktop) a visible window with the form still on screen.
\nStop the daemon, then verify it is gone:
\nopenbrowser-ai daemon stop\nopenbrowser-ai daemon status\ndaemon stop closes every tab, exits Chrome, flushes saved cookies/login state to the profile (so the next run reuses the login), and shuts down the daemon process. daemon status should report the daemon is not running. If it still reports running, the daemon is wedged, force-kill it:
pkill -f 'openbrowser.*daemon' || true\nForm runs can fail mid-workflow (validation error, CAPTCHA, network drop). Guarantee cleanup with a shell trap so a half-filled form never leaks a browser:
\ntrap 'openbrowser-ai daemon stop >/dev/null 2>&1 || true' EXIT\n# ... openbrowser-ai -c calls here ...\nDo not rely on the idle timeout. Do not call done() as a substitute, done() only marks the task complete inside the agent loop, it does not close the browser.