Choose Tidio Lyro when
Small teams that want live chat, a shared inbox and an AI agent from one vendor, with the model managed for them.
RAUM AI vs Tidio Lyro
Tidio Lyro or a website chatbot that runs on your own AI account?
Both answer visitors on your site from your own support content. The difference is where the model bill lands, and how much of the reasoning you can inspect before an answer is sent.
Last reviewed: September 2, 2026
14-day free trial. No credit card required.
Decision guide
Which approach fits best?
Choose RAUM AI when
Teams that already have documented procedures, want to pay their AI provider directly, and want to read the reasoning behind an answer.
Neutral considerations
What to validate before choosing
Packaged versus separate
Tidio bundles live chat, an inbox and the AI agent. RAUM Chat is an answer layer, so a team that also needs a human chat console should account for that separately.
Model ownership
Tidio states that it powers Lyro with Claude and its own in-house models. RAUM asks you to connect your own provider account, which is more setup and more control.
Early access
RAUM Chat publishes plans, but onboarding is currently run by our team rather than self-serve. Confirm timing before committing to a launch date.
Side by side
RAUM AI vs Tidio Lyro
Competitor cells use the official sources listed below. “Not publicly documented” means we did not find a current official statement; it is not a claim that the capability is absent.
| Decision area | Tidio Lyro | RAUM AI |
|---|---|---|
| Metering unit | Per Lyro conversation, defined as an interaction with at least one AI reply[1] | Per conversation, charged once per conversation id |
| Published entry price | Starter listed at $24.17/mo with 50 one-off Lyro conversations; Lyro volume sold as an add-on[1] | $49/mo for 500 conversations, $149 for 2,000, $399 for 6,000 |
| AI-provider credentials | Vendor-selected: Tidio states it uses Claude and in-house models[2] | You connect your own OpenAI, Anthropic or Google account |
| Model spend | Included in the vendor's per-conversation price | Billed by your provider at your rate, with no markup from us |
| Knowledge grounding | Scrapes your support content as its knowledge base[2] | Written procedures and help content you approve, plus connected read-only systems |
| Connected actions | Verify the current action catalogue against your systems | Read-only sources by default; state-changing actions need explicit approval |
| Live web knowledge | Confirm crawling, exact live-page access, and freshness behavior in the cited product documentation | Website crawling, exact live-page lookups, dynamic URL variants, and durable last-good snapshots |
| Workflow review controls | Confirm current builder, approval, review, and simulation controls in the cited documentation | Guided connected workflows, per-finding AI Review, ticket directives, and Simulation Mode |
| Zendesk onboarding | Product-specific installation and setup | Guided Marketplace installation with automatic webhook and trigger setup |
In depth
What the choice actually turns on
A bundled model price is a simpler bill and a fixed ceiling
When the vendor owns the model relationship, one number covers everything and you never think about tokens. The trade is that the cost curve is theirs: you cannot move to a cheaper model when one appears, or a stronger one when a hard ticket class needs it, without waiting for the vendor to offer it. RAUM takes the opposite position — you bring your own provider account, so the model bill arrives from your provider at whatever rate you have negotiated, and the subscription covers the platform rather than the inference. That is more setup on day one and more control every month after.
Ask what happens on the answer the bot should not give
Deflection rates describe the easy half of the queue. The half that decides whether you can leave a chatbot running unsupervised is what it does when the knowledge is missing, the visitor is not who they claim to be, or the request would change something. RAUM scores every answer's own confidence and lets you set which levels may reach a visitor at all; anything below the bar becomes a handoff rather than a message. Whatever you evaluate, test that path first — it is the one that costs you a customer when it is wrong.
Setup and migration
A practical evaluation sequence
- 1
Inventory the content Lyro is grounded on today — scraped pages and question-and-answer pairs — and decide which of it should become a written procedure rather than a page.
- 2
Open a provider account first. RAUM Chat answers with your credentials, so nothing runs until that connection exists.
- 3
Keep both widgets live on separate pages during the comparison and read the reasoning trail on the RAUM side before widening the origin allowlist.
FAQ
Questions teams ask
Does RAUM Chat include a live-chat inbox for human agents?
No. RAUM Chat is an answer layer, not an agent console — there is no seat to buy and no shared inbox. When a conversation needs a person it hands off, by email, a signed webhook, or by opening a Zendesk ticket with the transcript, with the visitor's explicit consent. If you need a human chat console as well, plan for that separately.
How is a conversation counted?
Once, per conversation, keyed on the conversation id — a long back-and-forth is one charge, not one per message. Included volumes are 500, 2,000 and 6,000 a month depending on plan, and anything beyond that is billed at $0.10, $0.07 or $0.05 per conversation.
Do I pay RAUM for the AI itself?
No. You connect your own provider account and that provider bills you directly at your own rate; we add no markup to model spend. The subscription pays for the platform.
Can I use it before I have written procedures?
You can, but it will not be good. RAUM answers from documented process rather than general model knowledge, so a team with real runbooks and a maintained help centre sees value immediately and a team without one mostly sees the model decline to answer.
Is RAUM Chat available to buy today?
The plans are published, but chat workspaces are currently provisioned by our team rather than through self-serve checkout. Talk to us about timing before you plan a launch around it.
Evidence
Official sources
- [1] Tidio pricing
Plan names, listed prices and the definition of a Lyro conversation; verify at purchase.
- [2] Lyro — AI Customer Service Agent for Better CX
Product naming, knowledge sourcing, and the statement that Lyro is powered by Claude and in-house models.
Product packaging and availability can change. Recheck these official pages and obtain a vendor quote before making a purchase decision.
Keep exploring
Related resources
Evaluate RAUM AI with your real Zendesk workflows
Install from the Marketplace, connect a supported provider, and verify representative tickets without sending customer replies.
14-day free trial. No credit card required.