Pricing, side by side
| Product | Free tier | How usage is metered | Paid |
|---|---|---|---|
| Flocci Chat | 500 free credits on signup (≈500 answered messages) | 1 credit = 1 answered visitor message. Nothing else is metered — crawling, uploads, tickets, analytics and unanswered turns are free | Platform Pro ₹499/mo or Pro+ ₹2,999/mo at account.flocci.in covers every Flocci app, not just Chat |
| Custom build (India, agency or in-house) | None — you pay from day one | Build cost, then LLM tokens, embedding calls, vector-database hosting, app hosting and engineering time for every change | Published Indian agency ranges run roughly ₹35,000 for a thin wrapper to ₹8,00,000+ for a production RAG system; running costs and maintenance are additional and permanent |
Competitor pricing was taken from the vendor's public pricing page on 29 July 2026 and may have changed since — always confirm on their site. Flocci Chat pricing is current.
What a support-grade RAG system actually contains
The demo is a weekend. Production is not. To be trusted on a real website an assistant needs: HTML extraction that survives messy markup, section-aware chunking, near-duplicate suppression, an embedding pipeline with re-index on change, hybrid vector plus full-text retrieval, diversity reranking so three chunks of the same paragraph don't fill the prompt, a confidence gate that decides answer versus clarify versus escalate, streaming with citations, a ticket path with transcripts and email replies, rate limiting and tenant isolation, and telemetry that tells you when quality drops. Every one of those is in Flocci Chat because every one of them had to be built to make the answers reliable.
The two costs teams forget
Refresh. A build that indexes your site once is stale in a month. Somebody has to own the re-crawl, the diff, the re-embedding and the deletions. Flocci Chat re-learns a page the next time a visitor opens it, keyed on a content hash. Failure behaviour. The interesting question isn't "can it answer the FAQ" — it's what happens on the question nobody wrote content for. Untuned RAG improvises. Flocci Chat's answer policy asks a clarifying question, or offers a human and opens a ticket with the transcript attached.
When building genuinely wins
If conversational AI is your product, or you have a hard requirement for a self-hosted model in your own VPC, or you need retrieval over data that cannot leave your network, build it — and Flocci Chat's approach is a reasonable blueprint to copy. A useful middle path: embed the widget now, learn from six months of real questions and the dashboard's unanswered-intent list, and let that evidence, rather than a guess, shape what you eventually build.
When a custom build is the better choice
- The assistant is the product you sell, not the support layer around it.
- Compliance requires the model and the vectors to stay inside your own infrastructure.
- You need retrieval over private systems that no external service may reach.
When Flocci Chat is the better choice
- Your site is React, Vue, Webflow or any JS-rendered stack. Flocci Chat's widget learns pages as visitors actually see them — after scripts run — where crawl-only trainers see an empty shell.
- You want a bill you can predict. One answered visitor message costs exactly one credit. No per-resolution judgment calls, no model multipliers, no seat fees.
- You're pricing in rupees. Credit packs and the ₹499/₹2,999 platform plans are India-first; one wallet covers the entire Flocci suite.
- You need answers you can defend. Every reply is grounded in retrieved content with sources shown; low confidence triggers a clarifying question or a human handoff, never a guess.
- You want tickets without another tool. Escalations land in a built-in dashboard inbox; operator replies reach the visitor in-chat and by email.