Teneo Protocol

Live access to the open web, built for AI.

[Teneo Protocol @teneo_protocol]

Today, we're launching the Live Web Data API by Teneo Protocol to the public.

One API that delivers public web pages live as clean Markdown or text, ready for AI agents and apps.

[Teneo Protocol @teneo_protocol]

Keep your Beacon running!

[phoenix @phoenix_jnr]

AI is only as good as the information it can reach. If it can't see what's on the web right now, its answers fall behind.

That's what @teneo_protocol fixes. People like me run an app called Beacon, and our devices help AI pull live info from the web.

10M+ requests went through the network in 4 days. That's about 30 every second, and it's the most it has ever handled. Watching it grow from the inside is wild.

[Teneo Protocol @teneo_protocol]

Up to 10 URLs per request, and up to 20 requests per second. Custom limits for large enterprises.

Docs, keys and pricing: https://t.co/OZ35BfejUl

[Teneo Protocol @teneo_protocol]

Today, we're launching the Live Web Data API by Teneo Protocol to the public.

One API that delivers public web pages live as clean Markdown or text, ready for AI agents and apps.

[Teneo Protocol @teneo_protocol]

Built for AI agents and RAG pipelines, research and investment teams, legal and compliance, SEO and AI-visibility teams, and news and brand monitoring: anyone who needs public pages as they stand today.

Use it over REST or MCP, with one key.

[Teneo Protocol @teneo_protocol]

@ZellaSama Hi Zaykon, please contact Support on Discord

[Teneo Protocol @teneo_protocol]

We compared old copies of public web pages with what those pages say now. A third had lost at least 5% of their text within 30 days.

A RAG index is a copy, and a copy stops changing the day you make it.

We took a fixed list of pricing pages, documentation and official rates, and compared Internet Archive captures of different ages.

After 30 days, 16 of 48 pages had lost at least 5% of their text. After 180 days, 31 of 47 had. After 30 days, 9 of 48 had lost at least one price, percentage or version number.

Mistral's pricing page listed 10, 15 and 30 dollars a month in API credits on 7 October. By 12 October it listed 15 and 25.5. A copy indexed on 7 October still says 10 and 30, with a valid citation and no error.

The usual fix is a webhook, and that only works for sources you own. A page on someone else's website has no webhook for you. So you have to go and look.

27 of the 54 live pages we tested can be checked without a download, because they answer a conditional request with a 304. The rest have to be fetched and compared.

Teneo supplies that re-fetch: the page requested from the market you name, through a connection whose owner agreed to share it, paid per request.

It is a small hand-picked list, so treat it as a floor. The post covers how to spot changes on a site you do not own, how often to re-fetch each kind of page, and the free seven-day pilot for RAG teams:
https://t.co/lbEB7l790w

Building RAG on public pages?

Reply with the sources you trust least and the markets they serve.

[Teneo Protocol @teneo_protocol]

@KreatValue Haha, right? More information on TGE soon

[Teneo Protocol @teneo_protocol]

@DigsdotHQ Thanks, Digs

[Teneo Protocol @teneo_protocol]

@dair_ai The contract before the agent starts is a good idea. Could the same thing work for agents doing web research, where you lock in what sources and evidence count before they begin?

[Teneo Protocol @teneo_protocol]

@Yuchenj_UW Security gets a lot harder when you don't know where the data is coming from or who has access to the infrastructure. How much of your security review is about the data providers themselves?

[Teneo Protocol @teneo_protocol]

@RimonR23 The shared logbook idea is great. I’d want every saved piece of information to keep its original source and date too, so agents know what’s still worth trusting.

[Teneo Protocol @teneo_protocol]

@dair_ai Could the same approach work for agents that spend most of their time fetching web data? Cutting repeated source reads could probably save a lot without changing the model.

[Teneo Protocol @teneo_protocol]

@Dr_Singularity A living database would be useful, especially if agents could keep it updated as new discoveries come out. Who verifies each entry though?

[Teneo Protocol @teneo_protocol]

@omarsar0 The hard part will be keeping that graph fresh. How does it handle a document or source changing after the agent has already built its memory?

[Teneo Protocol @teneo_protocol]

@mardehaym I like the idea of pushing repeatable work into code. For the parts that still need the model, I'd want every external source it used recorded so you can check the result later.