Answer every visitor.
Keep every lead.

Bot chat runs on your own LLM key. Rylvo adds no markup and charges nothing per message. Fair use applies; see Bot replies.

Automated safety rules enforced before a response is sent.

Customers can send photos, screenshots, documents and voice notes on any channel — the bot reads and hears them. Runs on your own LLM key, so there is no markup.

Photos, screenshots and document pages processed per billing period.

Audio transcribed per billing period. Rounded up per clip.

Docs, files, and URLs your bots retrieve answers from.

External tools and APIs your bots can call.

Deterministic step-by-step conversation paths the engine enforces — per-step goals, required data, allowed tools.

Set your own pass bar, weigh what matters, add score dimensions of your own, and see which tests would flip before you save.

Check a reply for a pattern, JSON of a given shape or a number, and check what the bot did: the tools it called, whether it used the knowledge base, how fast it replied and what it cost. Each check fails the test on its own — no AI judge can overrule it.

Your reviewers set their own verdict on judged test results; see how often the AI judge agrees with them (with Cohen's kappa and a confusion matrix), where it is too generous or too strict, and whether a scoring change makes it agree more.

Hold a release to Production until its tests clear the bar you set — a minimum pass rate, no failing critical test, a named test set, how often the judge agrees with your reviewers — and ask the same gate from your own CI pipeline.

Every save of your detection rules, test scoring and release gate is kept as a version on every plan. Put an earlier one back in a click — as a new version, checked like any save — and export or import them as JSON files.

Lock a detection rule, the test pass bars or a score dimension so every bot follows the workspace's setting. A bot's own change to it is kept but not applied, and a lock already saved keeps holding if you change plans.

Edge Cases, Agent Evolution, Prompt Optimization, Red Team and Broadcasts run on your own LLM key — free, zero markup. No AI credits, no markup. Knowledge-base indexing is included up to your plan's allowance. The only wallet pays for storage overage and optional add-ons.

Reviewer agents analyze your runs and auto-suggest guardrails.

Instructions a bot follows only when a customer's message calls for them — with a library of ready-made ones, versions and automatic rollback.

Suggests prompt changes for you to review (OPRO, APE, DSPy and PromptBreeder strategies).

Finds failure patterns in production and auto-generates guardrails.

Every automatic scan also reads recent conversations against your AI review criteria, on your own key, under a cost cap you set.

Rules you write over what a conversation records — file a case when the bot promises a refund, names a competitor or takes too long. Tried on recent conversations before you save.

Adversarial inputs that probe your bot for weaknesses.

Push every conversation to your own database or webhook.

Conversation storage events included free. Overage: $0.50 per 1K events.

How long traces are kept. Extend with retention add-ons.

Curated collections of traces for analysis and review.

Distinct traces protected from auto-deletion across all datasets.

Step through any past conversation, turn by turn.

Rooms of charts on every reply — volume, quality, speed and cost — plus alerts when a number crosses a line.

Real-time oversight of live conversations with intervention.

Cron-based scans, reviews, and report generation.

Slack, email, webhook, Discord, Teams.

Send messages to your audience across WhatsApp, Telegram, Email, SMS, and more.

Free and Lite are single-seat plans. Pro starts at your seat and adds teammates for $40/seat/month ($32 on annual) — prorated instantly, no plan change.

Pro/Enterprise on both sides. The agency sends a role-scoped request; the client owner approves it. Client ownership, data, settings, and billing remain isolated.

Replies your bots send, on your own AI key with zero Rylvo markup. Fair use: there is no per-reply charge. Pro's guide figure is about 100,000 replies a month; Enterprise volume is set in your contract.

Enterprise: named contact, next business day (not an SLA). Other paid plans: email, no response target. Free: docs.