Tier-1 SaaS support, answered in two seconds.
Account questions, password resets, billing, plan changes, and onboarding pile up faster than five engineers can clear. We answer the repetitive 60 percent with citations and hand tier-2 to your team with full context.
Most of your inbox is the same five questions on repeat.
Password resets, seat changes, invoice resends, plan upgrades, SSO setup. The five questions that consume an engineer's afternoon are the five the AI answers in seconds, with a citation from your docs, in the customer's language.
- Reads your docs site, help-centre, and pricing page to answer policy and product questions.
- Hands tier-2 (debug logs, custom configs) to your engineers with a clean summary.
- Replies in the customer's language, EN and NL are most polished, plus 100+ more.
- EU-hosted, GDPR-first, DPA signed per workspace, every reply auditable.
The widget on your app and marketing site, the email channel on support@. The AI handles tier-1 in seconds, hands off tier-2 with context, and lets your engineers ship instead of triage.
Built for the support volume of a real SaaS team.
Six things that change in the first week of going live, in the order your team feels them.
Knowledge from your docs, in seconds
Paste your docs URL, your help-centre URL, your pricing page. The scanner reads them once, indexes them, and every AI reply cites the page it came from. Re-crawls run on a schedule so doc updates land in the AI within hours.
Smart escalation to the right engineer
Add specialties to each teammate, billing, integrations, SSO, performance. When the AI hands off, it routes to the operator whose specialties best match the topic. Engineers see only conversations meant for them, not a firehose.
Confidence floor, no bluffing
Every reply gets a confidence score from retrieval quality, citation count, and language-model uncertainty. Below the threshold (default 70 percent), the AI stops answering and hands off. Customers never get a confident-sounding wrong answer.
Multilingual on day one
Auto-detected from the first message, locked for the conversation. Dutch and English are the most polished. Add Q&A pairs in any language and the AI uses them. Useful if you sell across the Benelux and DACH from one inbox.
Suggest-only mode for sensitive tickets
Toggle auto-send off and the AI writes a complete draft with citations. A human clicks send. Useful for the first two weeks, for legal-adjacent questions, or for any team that wants a check before a customer-facing reply.
Custom HTTP for your own backend
Wire your subscription state, usage data, or feature flags through a custom HTTP tool. The AI can answer questions like, is my plan over its seat limit, when is my next renewal, did my invoice go through. Configure once, the AI calls it like any other function.
Common questions from SaaS support teams.
Will the AI hand off complex technical tickets, or try to answer them?
It hands them off. The confidence floor (default 70 percent) catches anything where retrieval doesn't ground a clear answer. Debug logs, custom API errors, anything edge-case-shaped lands in your engineers' queue with a summary, the original message, and the AI's best guess as a starting point. Your team picks up at minute one, not minute fifteen.
Does Keloa work with our existing helpdesk like Zendesk or Intercom?
Keloa replaces them. The Keloa inbox handles chat and email conversations end-to-end. Most SaaS teams migrate fully because the per-seat math no longer makes sense once the AI handles 60 percent of volume. We can migrate your existing knowledge base in a day. If you want to run side-by-side for a few weeks, set Keloa on a subset of inbound and compare deflection.
How does Keloa handle account-specific questions like billing or seat usage?
Through a custom HTTP tool. Configure your billing API endpoint or seat-usage endpoint once, with the customer's identifier as a parameter, and the AI can answer concrete account questions in real time. Without that wire-up, the AI sticks to policy answers from your docs and hands account-specific questions to a human.
Which plan fits a typical 10-seat SaaS support team?
Business at 149 euro per month covers most teams, 6,000 AI replies, five seats, SLA timers, proactive triggers, advanced analytics. If you need more than five seats or you push past 6,000 AI replies a month, top-ups are 30 euro per 1,000 extra replies. Heavy volume (over 25,000 replies) moves to Scale at 449 euro per month with 25 seats included.
Can the AI write code snippets or help with API integration questions?
If your docs cover it, yes. The AI pulls the example from your developer docs and cites the page. For original code or debugging of a customer's specific setup, the AI hands off to an engineer, code generation outside the bounds of cited documentation is the kind of place where confidence drops and handoff fires.
Is the AI safe for compliance-heavy customers (GDPR, SOC 2, HIPAA-adjacent)?
EU-hosted in Amsterdam, GDPR-first, DPA signed per workspace, every reply auditable with a citation back to a source. PII is redacted from outbound payloads to the language model. Sub-processors are listed publicly. ISO 27001 is on the roadmap. For HIPAA-specific deployments, talk to sales, we'll be honest about what's in scope and what isn't.
Run your SaaS support like a calmer team.
Free Starter (50 replies), no credit card. Paste your URL, paste your docs, watch the AI answer your top tier-1 questions in your tone within ten minutes.