New — the Tender Checklist Workbench is live: review tenders with page-cited answers. See it →
← All products

RapidCloud Chat

One grounded conversation across all your agents

RapidCloud Chat is the central AI interface for your private platform, hosted in the EU. It grounds every answer in your own documents and data and returns inline citations so you can trace each claim to its source. From one conversation it reaches your research, infrastructure, residue, and files agents, and it can find images by meaning or visual similarity. Answers can render the matching images inline alongside cited text.

Every agent. One conversation. Grounded in your data.

RapidCloud Chat is the front door to your private AI platform. Ask a question and get an answer drawn from your own documents — with inline citations you can follow to the source. Reach your research, infrastructure, residue, and files agents without leaving the thread. Search your images by meaning or visual similarity, and see them rendered inline beside the text that explains them.

It’s a dark-first premium interface that’s just as sharp on a phone as on a desk, with per-conversation privacy modes and single sign-on built in. EU-hosted, access-gated, and connected to everything you already run on RapidCloud.

→ Start a grounded conversation at chat.rapidcloud.eu

Under the hood▾

Grounding & citations. Chat runs retrieval over your corpus and returns inline citations linked to source documents, so every claim is traceable rather than asserted. Tool-calling agents. A multi-agent runtime exposes your research, infrastructure, residue, and files agents as callable tools in a single thread, routing each turn to the right backend. Visual search. CLIP-based image_find does text-to-image semantic search and image_similar does visual nearest-neighbor lookup against the imagestore; results render inline via standard markdown image syntax and populate a side panel, filtered to images the model actually used. Model routing. Per-conversation modes (private local model, shared org key, or bring-your-own-key) are stored on thread metadata; o-series/gpt-5 requests are normalized (max_completion_tokens, no temperature) before dispatch through a LiteLLM gateway. Security & EU hosting. Self-hosted on the RapidCloud cluster, fronted by Cloudflare Access with Entra ID SSO and strict CORS; auth is enforced by hostname over HTTPS, and your data stays on EU infrastructure under your control.