Taranis AI

Free

Taranis AI does not publish a subscription on the pages opened on 17 Oct 2026. The GitHub license field is EUPL-1.2, the homepage says the European Union Public Licence, and the copyright line is © 2025 AIT. The software is what you deploy with the Docker Compose guide. The project does not host a public newsroom for your queries. There is no seat price and no usage meter on the homepage or on release 1.5.1. Models you connect for the RAG chat may cost money somewhere else. The Taranis docs reviewed here do not print that model fee. Budget for the host, Redis, and the sources you are allowed to collect.

The Taranis homepage and the README describe the same pipeline. The project collects unstructured news from websites, enriches it, and lets an analyst turn articles into a report that can become a PDF. Features named on the homepage are a chat over collected stories using retrieval-augmented generation, a task queue via RQ and Redis, NLP enrichment and story clustering, entity recognition for cybersecurity terms such as APTs, CVEs, and IoCs, and advanced search by tags and sources. The homepage says the workflow from earlier Taranis versions is integrated, and that the project was inspired by Taranis3 and Taranis-NG.

• EUPL-1.2, self-hosted. The API license field matches the homepage's European Union Public Licence line. You run the Compose stack.
• Release 1.5.1. Published 9 Oct 2026. The API push on the review date was 15 Oct 2026.
• Collect, then edit. The homepage describes RAG chat, an RQ queue, clustering, and entity tags. Analysts still write the report.

1. 1

   Follow the Docker Compose deployment guide for a production install. The README points at that page.

2. 2

   Cite release 1.5.1, published 9 Oct 2026, until you have a newer tag.

3. 3

   Open the source story behind a chat answer or an entity tag before you brief it.

4. 4

   Read the EUPL before you redistribute a modified copy.

• A team wants news monitoring without a paid intelligence seat. Taranis AI is the stack you deploy. The homepage describes collection from websites, a RAG chat over stories, an RQ and Redis queue, NLP enrichment, story clustering, and entity recognition for terms such as APTs, CVEs, and IoCs. The Docker Compose guide is the production path the README links. Release 1.5.1 was published 9 Oct 2026. The API license field that review day was EUPL-1.2, matching the homepage's European Union Public Licence line and the © 2025 AIT notice.
• A chat answer is pasted into a finished report. The homepage says analysts refine AI-augmented articles into structured reports. Open the story the answer used, and keep the entity tag attached to that story rather than to a conclusion. The project was inspired by Taranis3 and Taranis-NG. Those names are history. They are not this release. The README's marketing line about changing collection is not a measured result and does not belong in the finding.
• Someone cites 1,230 GitHub stars as a quality score. That was the API count on 17 Oct 2026, with 181 forks, on the master branch, last pushed 15 Oct 2026. It will change. The homepage does not print a price. Models you attach for the chat may cost money at the model provider. The Taranis pages reviewed here do not list that fee. One screenshot in the repository is not a tour, which is why this listing omits the Screenshots section.

Hard cases

The case where the first result is easy to over-read.

Questions

Titled sources. These public pages are what the quotes on this listing come from.