Notte

The Web Automation Platform

[Notte @nottecore]

Anything API is live on Product Hunt! 🔥🚀

Most websites don't have public APIs. Anything API fills that gap.

Describe the browser work you need. Our agent builds it, deploys it, and hands you a callable endpoint that you or Claude can invoke from everywhere.

Any website. We deliver the API.
https://t.co/KFobEjRpaw

[Notte @nottecore]

Agentic users of @NousResearch, @nanobot_project,
@openclaw, and more are now welcome to automate behind login!

Simply BYO Agent. Pre-packed:

✓ Hosted browser sessions
✓ Isolated credentials Vault for auth
✓ Agent swap? Flow stays the same

The setup is one prompt. Enjoy!

[Sam Morris @samatnotte]

becoming a daily routine.
not even looking at variations in latency just the UI 🌸

[Notte @nottecore]

Agentic users of @NousResearch, @openclaw, @nanobot_project, and more are now welcome to automate behind login!

Simply BYO Agent. Pre-packed:

- Hosted browser sessions
- Isolated credentials Vault for auth
- Agent swap? Flow stays the same

The setup is one prompt. Enjoy!

[Notte @nottecore]

Browser Arena now runs daily 🌤️

Track every major cloud browser provider.
Watch latency and reliability drift over time.

• Same infrastructure
• Full historical data included
• Now automatically benchmarked

Open-source.
Reproduce, contribute, and run it yourself.

[Notte @nottecore]

▸ Console: https://t.co/zY4zeXbXDU
▸ Get Setup: https://t.co/ufkKJYFGuR

[Notte @nottecore]

Agentic users of @NousResearch, @nanobot_project,
@openclaw, and more are now welcome to automate behind login!

Simply BYO Agent. Pre-packed:

✓ Hosted browser sessions
✓ Isolated credentials Vault for auth
✓ Agent swap? Flow stays the same

The setup is one prompt. Enjoy!

[Notte @nottecore]

https://t.co/UydEuy1J7u

[Notte @nottecore]

Browser Arena now runs daily 🌤️

Track every major cloud browser provider.
Watch latency and reliability drift over time.

• Same infrastructure
• Full historical data included
• Now automatically benchmarked

Open-source.
Reproduce, contribute, and run it yourself.

[Andrea Pinto @ogandreakiro]

Congrats on the $800M! 4x deserved for this team.

[Notte @nottecore]

We partnered with @FireworksAI_HQ and ran 720 browser agent tasks across 4 LLMs.

The model that looks like a bargain on token price was 2.3x more expensive per successful task than the winner.

Here's what was happening underneath.

[Notte @nottecore]

Two key takeaways:

• Tail-latency consistency across heterogeneous architectures is a serving-layer property that @FireworksAI_HQ's infra makes visible.
• Parse retry rate should be a first-class metric in every agent benchmark.

[Fireworks AI @FireworksAI_HQ]

We ran 720 browser agent tasks with @nottecore across frontier models.

One baseline model produced malformed outputs in ~1 out of every 5 calls, leading to retries inside multi-step workflows.

Across Kimi K2.5, GLM-5, and MiniMax M2.5 served on Fireworks, retry rates were near zero and latency stayed stable even as tasks extended across multiple steps.

Same workload. Same agent loop. Different execution behavior.

That gap is what shows up as cost, latency, and reliability divergence in production agent systems.

Read the report: https://t.co/6thZVvLomR

[Sam Morris @samatnotte]

built a free site that uses real market data + agentic research to tell you if your company is cooked.

apologies in advance.

[Notte @nottecore]

Across three models with very different architectures, p95/p50 latency stayed inside a tight 1.9-2.3x spread.

Tail-latency consistency across heterogeneous architectures is a serving-layer property @FireworksAI_HQ's infra makes visible.

▸ Full report: https://t.co/Nsn2dYKkal

[Notte @nottecore]

We partnered with @FireworksAI_HQ and ran 720 browser agent tasks across 4 LLMs.

The model that looks like a bargain on token price was 2.3x more expensive per successful task than the winner.

Here's what was happening underneath.

[Notte @nottecore]

A note on method: we ran text-only to keep the four-model comparison fair (only 1 of 3 Fireworks models supports vision). That handicaps Gemini on its strongest dimension.

However we ran a vision check anyway: it gains +6pp, and MiniMax M2.5 text-only still beats it.

[Notte @nottecore]

Reliability, not intelligence. Why?

One model wasted nearly 1 in 5 LLM calls on malformed JSON it had to retry. Each retry burns a full inference call's worth of context and 2-3s of latency.

Every static benchmark misses this because it happens inside the engine.

[Notte @nottecore]

@Shpigford @browserbase 1. ✓
2. ✓
3. ✓ -- secure, shareable session links users can watch live and download as MP4s: @getdiana's already been shipping to their customers https://t.co/LSRF6BZAG3

happy to help you get started with some credits

[Notte @nottecore]

Diana, a Slack-based AI assistant platform that connects employees to thousands of tools from a single interface, used Notte to give their AI agents:

• Browser access to tools without APIs, using secure user-authenticated sessions
• Live session visibility through a secure, shareable link, so users can watch what the agent is doing in real time

"The live session viewer isn't just a technical feature but a transparency guarantee. This is a significant matter when you're asking someone to let an AI act on their behalf." — @upekabee, CEO of @getdiana