Demonstrate by Notte - Browser workflows to deployed automation in minutes
by•
Record any browser task once and get production-ready code instantly with Demonstrate Mode.
Edit further your code in our Automation Studio with live browsers, deploy automation code as a serverless function, and schedule it to run autonomously. Managed sessions, proxies, identities, and vaults handle everything behind the scenes.
The fastest path from prototype to production in one unified platform.
Replies
@ogandreakiro andrea_pinto Love the hybrid “Demonstrate → editable script” + Agent Mode; at scale the hard part is selector/DOM drift + auth/2FA fragility that causes silent flake across thousands of runs.
Best-practice: generate resilient locators (ARIA/role/text + fallback strategies), snapshot+diff the DOM, run canaries, and enforce idempotency/anti-dup guards (especially for form submits) with durable retries.
Open Q: how do you version/replay sessions for debugging (video/trace + DOM snapshots), and can vault + identity access be scoped per workflow/tenant with audit logs? 🔥
Notte
@ryan_thill hey Ryan! Yes definitely - selector and DOM drift is a major challenge we're trying to address with resilient self-healing functions. We have version control on your deployed functions (every deployment is one version) and replays for any sessions you run through Notte is always available to you. You can scope vaults and agent identities to function tenants as well.
@ogandreakiro Love that you’ve got versioned deploys + replayable sessions baked in, that’s the foundation.
Curious how your “self-healing” works in practice: are you generating locator sets (role/aria/text + structural anchors) and scoring candidates from DOM diffs, or learning per-site over time?
Also for replays, do you persist a deterministic artifact bundle (trace/video + DOM snapshots + network/console) with stable IDs so a run is fully reproducible even after the target UI changes? And for vault/identity scoping, is it enforced via per-tenant KMS keys + audit trails on secret access?
Looks nice. BTW, how it works under the hood? does it use some kind of Playwright based automation or headless browser?
Notte
@atairov Hey Tairov, thanks! Demonstrate Mode runs on a cloud browser we operate as part of our home rolled infrastructure (PS: You could use these sessions as standard browser as a service API if needed). Everything done in Demonstrate sessions is recorded and ported in an automation function expressed with Notte Framework - our 100% Playwright compatible built for browser agents and AI automations.
Yapify
congrats on the launch!
Notte
@dan_meier1 thanks Dan! let us know if you get a chance to try it out
Pane
Notte
@rohan_bajpai1 Thanks Rohan!
Notte
@rohan_bajpai1 glad you see the vision:)
Great product, need to give it a shot.
Notte
@jeetendra_kumar2 Thanks!
Notte
@jeetendra_kumar2 thanks - let us know what you think!
panda{·}etl (YC W24)
@ogandreakiro congrats on the launch
Notte
@gdc Thanks Giuseppe!
Demonstrate Mode helps bridge the gap between trying something out and actually using it in production. For me, recording a task once and getting real code I can edit solves a problem I keep running into with other automation tools. Creating the automation, testing it in a live browser and deploying it from the same place keeps everything in one flow. That last part is usually what’s missing when I try to move browser automations into production.
Notte
@henrik_falk that 'last part' was a problem we kept running into, so we built our console and Demonstarte Mode around the idea of going from prototype to production in one environment - glad you like it!
Trace
Notte
@timcha_cherkasov 🔥🔥🔥
does the generated code output standard Playwright/Puppeteer that we can tweak manually, or is it a proprietary syntax?
I wasted an embarrassing amount of time just trying to get a basic browser automation to run. One tool to record another to fix things then something else to actually execute it. Half the time it just dumb.
Tried Demonstrate Mode out of curiosity. Showing the flow once and getting code right away is the first time this hasn’t felt overengineered. Still early but not having to bounce between tools already makes it worth paying attention to.
Best of luck team with the launch.