PropensityAI

What PropensityAI is

PropensityAI combines retrieval and reasoning to produce concise, cited search results from the web, documents, and media. It links important claims to sources and can indicate uncertainty or decline to answer when supporting evidence is lacking. Users can query research papers across PDFs and citations, then compare methods, results, and limitations. Image analysis includes OCR, captioning, visual question answering, and structured extraction. Audio analysis can transcribe, identify speakers, summarize, and extract tasks; video search can locate scenes and spoken content with time-range citations.

The service lists source filters and typical internal knowledge-base connectors for shared drives, wikis, ticketing systems, and internal documents. Its free Starter plan requires no sign-up and allows up to 16 sources per answer and one file per query. Team and Enterprise add shared-document or governance capabilities, while Enterprise supports on-premises or VPC deployment. The maker also offers local LLM systems that can run in customer infrastructure, including air-gapped deployments; its local packages are one-time purchases with no per-token usage fees. The site says private queries and uploaded files are not used to train foundation models. PropensityAI ships from Berlin, Germany, and its privacy policy names FetchCFD UG (haftungsbeschränkt) as the responsible party.

Who it suits

PropensityAI suits people who need cited search across web sources, documents, media, or research PDFs. Team is described for groups working with shared documents and connectors; Enterprise is aimed at secure deployments, VPC, and governance needs.

Limits

Starter is limited to 16 sources per answer and one file per query, with basic citations. Its Standard and Reasoning models are best-effort and may throttle at peak; Team allows up to 30 sources and 10 files per query, while Enterprise allows up to 40 sources and 10 files per query.

Getting started

1. Open the PropensityAI website.
2. Use Starter without signing up, or choose a paid offering.
3. Submit a question and, on Starter, include no more than one file per query.
4. For shared documents and connectors, consider the Team plan.
5. For on-premises or VPC deployment, consider Enterprise or ask about local LLM systems.

Verdict

Pick PropensityAI if you want cited answers across mixed source types, with a no-sign-up free tier or local deployment options. Look elsewhere if the Starter limits are too restrictive and you need a published price for Team or Enterprise.