ROI Updates | 1 September 2026

 

Introducing Kubro Skills

Use Claude, ChatGPT, Gemini, or any LLM that supports Skills,

to manage Kubro and ground your research 

 

Research grounded in your targeted and filtered Kubro content should increase the relevance and accuracy of the answers you get. Now, you can ask your LLM app to use Kubro Collections.

 

Setting up 20 or 50 new sources and a daily scheduler in Kubro has always been a human job. From today, it can be a sentence typed into Claude.

 

A New UI For Kubro

 

Kubro can now be managed and used directly from Claude, ChatGPT, Gemini, or other apps and harnesses that support Skills. People have been saying "AI is the new UI" since about 2017. This is us building it for Kubro.

 

Two Skills are available today: Web Query Admin and Content Explorer. Observability (Health Check) will follow shortly.

 

Example: Set Up Web Queries And Schedulers

 

You want to start tracking data centre companies - projects, deals, customers, executive changes, reports that they disclose. 

 

You ask your assistant to find the pages where those companies publish, then set up the Web Queries and Schedulers in Kubro and get them running.

 

Using Claude As An Example:

 

 

 

Worth watching what happens before anything is written. Claude confirms the crawler type and whether the query should be enterprise or private. Then it flags that the sources landed in a query called "DC2", while the output collection is named "DC Majors", so you don't trip over it in the UI later. It asks rather than guesses, and it tells you what it did:

 

 

Set up by hand, that sequence is 20 URL lookups, so about 20-30 minutes in the Web Query admin. With Claude (in this example), it's a few sentences and about 90 seconds of your input time.

 

The Same Configuration in Kubro UI:

 

 

Example: Research Grounded In Your Own Filtered Sources

 

You want to check the latest property deals in the US REITs sector, answered from the content you have already built up from company announcements gathered in your Kubro account.

 

You ask, and the assistant pulls the relevant material from Kubro as context.

 

Using ChatGPT As An Example:

 

 

 

Grounded in your sources, not the open web. This is the difference between asking an assistant a question and asking it through Kubro.

On its own, it searches the open web and answers from whatever it happens to find. Through the Content Explorer Skill, it answers from the sources you selected, crawled, tagged and quality-checked, with the underlying articles and filings sitting right there to check. 

 

For analysts doing the data collection work, researching collections this way could be the first step to enrich and cross-check the data, given all the relevant content captured and filtered in Kubro already, before 'Googling' or open-ended LLM chats.

 

The Corresponding Collection 'Manual' View In Kubro UI:

 

 

A Note On Security

 

Kubro Skills don't store user credentials. You supply your Kubro API key in the chat session, and it is used only for that session. No persistent token is held by us or by the LLM provider, and revoking the key in Kubro cuts off access immediately.

 

Each Skill is scoped to specific Kubro APIs and cannot reach beyond them. The Content Explorer Skill is read-only. Actions taken through the Skill are all visible in the account, and Web Queries created through a Skill respect the same enterprise and private visibility settings as the UI, and the Skill does not have a 'delete' capability.

 

Content passed to the assistant is governed by the user's own agreement with Anthropic, OpenAI or other providers. Enterprise and business tiers on all three should exclude that content from model training.

 

 

About 'Skills' - If You Are Less Familiar

 

Skills are a set of detailed instructions for an assistant, plus permission to call certain Kubro APIs, so it can decide and execute specific actions in Kubro from natural language. Within boundaries we define, the assistant does the many manual steps for you.

 

Anthropic's introduction to Skills is here, and the open specification is at agentskills.io.

 

Kubro Skills build on the existing Kubro APIs: Web Query API, Search APIs (Web, Filings, Social, Email, Documents, Agents), Write APIs (Web) and AIDE (AI Data Extraction) APIs, and Observability (Health Check) APIs coming soon.

 

 

 

Access

 

The Web Query Admin Skill and the Content Explorer Skill are available to clients now at no additional cost. The Observability (Health Check) Skill follows shortly, on the same terms.

 

If you don't have Kubro access and would like to explore the use cases and applications with us, please let us know by replying to this email. We have a formula for setting up and running quick pilots, customised to specific use cases.

 

 

 

 

Note: Claude, ChatGPT and Gemini are trademarks of their respective owners. Kubro Skills is an independent integration and is not affiliated with or endorsed by Anthropic, OpenAI or Google.