CC-RAGOS turns your documentation — PDFs, tables, images, help articles — into cited AI answers, embeddable help centers and a full support desk. Self-hosted, no per-agent fees.
Fetched from the production system on this page load — not a case study, not a projection.
vs. answering everything manually
from real client users, thumbs on every answer
support tickets + AI-answered questions
same team, entire platform vs seat fees alone
SaaS fees avoided — — agent seats at $0/agent, forever
Screen recordings of the deployed products — real data, real answers. Click to play, fullscreen for detail.
No Dify, no black-box vendor. A FastAPI orchestration layer runs retrieval and ingestion; Next.js presents the evidence; Qdrant holds the vectors.
Docling handles scanned PDFs, complex tables and figures other pipelines choke on.
Agentic chunking with contextual retrieval — chunks keep their document meaning.
Embeddings land in Qdrant Cloud — one source of truth, hybrid dense + sparse search.
Multi-hop agentic retrieval filters, re-ranks and keeps only what supports an answer.
Streaming responses cite the exact chunks used — provenance in the UI, not a black box.
Feedback, doc-gap logging and a weekly digest turn every miss into new documentation.
A built-in MCP server exposes the live support and docs data as read-only report tools. Point any MCP client at it — Claude, Cursor, whatever your team uses — and compose custom reports in plain English. No BI tool, no exports, no waiting on an analyst.
Remote MCP server (Streamable HTTP), token-gated, read-only. The same data behind the desk dashboard and this page — queryable by any agent you trust.
Because seat fees scale with your team, and their AI can't cite your docs the way your own pipeline can. This category is real money — Salesforce paid $3.6B for Intercom's Fin AI agent in June 2026. We run the same capability in-house for about $240 a year.
| Zendesk / Intercom | Apache Answer | CC-RAGOS | |
|---|---|---|---|
| Cost per agent | $55–115 / mo | free | $0 |
| Cited AI answers | $0.99 / resolution (Fin) | none | included |
| Your infra, your data | their cloud | self-hosted | self-hosted |
| Multimodal ingestion | — | — | PDF · tables · images |
| Doc-gap detection | — | — | automatic |
| Running cost | $5,280+/yr (8 seats) | hosting only | ~$240/yr total |
A self-hosted, explainable multimodal RAG platform: one FastAPI + Next.js + Qdrant engine powering grounded doc-chat, embeddable help centers and a full support desk — deployed today for two client portals.
Real. Two client support portals run on it in production, and the numbers on this page are fetched live from the running system. AI support is a $2B+ market growing fast — this is the self-hosted end of it, running for real users today.
Free-tier cloud (Vercel + Qdrant + KV) plus roughly $20/month of LLM API usage. No per-agent seat fees at any team size.
Every response cites the exact source chunks it was grounded on, users rate every answer, and unanswerable questions are logged as documentation gaps with suggested titles.
That's the point. Docling-powered ingestion parses scanned PDFs, complex tables and embedded figures — the documents most pipelines skip.
One-click sync pulls an entire Answer-style Q&A board into the corpus, and the help center stays live-synced to it afterwards.
See the live system, the numbers behind it, and what it would save your team.
Watch it work