HomeBlog › letchat.ai: the customer chatbot you own — logic in code, data yours
Blog

letchat.ai: the customer chatbot you own — logic in code, data yours

Jul 19, 20265 min readBy the Rysh team

An embeddable support chatbot with knowledge grounding, lead capture, and human takeover — where the logic and the data stay yours.

If you run a small company, the website-chatbot decision usually looks like a two-item menu, and both items are bad:

  1. The big support suites. Excellent, mature — and priced for enterprises. Per-resolution or per-seat fees that scale against you, a black-box bot you can't inspect, and your entire customer-conversation history living in someone else's product. Leaving later means leaving the data.
  2. DIY. Wire a widget to an LLM API over a weekend. The demo is great. Then you discover the actual product is everything around the demo: grounding it in your docs so it stops inventing features, capturing leads, handing off to a human when it's out of its depth, analytics, keeping it on-brand — and now you maintain a chatbot platform instead of your business.

letchat.ai is the third menu item: a customer chatbot that behaves like the polished SaaS option, with the ownership model of the DIY option.

What you get

An embeddable widget, in minutes. A small snippet on your site — your colors, your tone, your domain. Custom domains are supported, so the chat experience is your brand end to end, not an ad for the chatbot vendor. No engineering project, no procurement cycle.

Answers grounded in your knowledge. The bot answers from the knowledge you give it — your docs, your FAQs, your policies. That's the difference between a chatbot that represents your business and one that improvises about it. And when the answer isn't in its knowledge, the right behavior is built in: say so, and offer a human.

Human takeover, built in — not bolted on. This is the feature that makes an AI bot safe to put in front of customers. The moment a conversation needs a person — the customer asks, the bot's out of depth, or your team just wants to step in — a human takes over the live chat and the bot steps aside. Your customers are never trapped in a loop arguing with software.

Lead capture. A visitor asking pricing questions at 11pm is a lead, not a ticket. letchat captures the contact and the context, so the conversation your bot had becomes the conversation your team continues.

Analytics and CSAT. What are people asking? Where does the bot fall short? How do customers rate the answers? You see the whole picture — conversation volumes, satisfaction scores — so improving the bot is a data decision, not a guess.

Teams. Multiple people managing conversations, sharing the takeover queue, working the same inbox. It's a support tool, not a solo toy.

The part that's actually different: ownership

Every feature above exists somewhere in the big suites. Here's what doesn't.

The logic is yours, in code. letchat is built on rysh, an agent platform whose core idea is that an agent's definition is a readable, versionable artifact — not a pile of toggles in a vendor dashboard. Your bot's persona, rules, and knowledge are things you can inspect, edit, and keep. When you change how the bot behaves, that change is explicit and reviewable.

The data is yours. Customer conversations are the most sensitive asset a support tool touches. letchat's model is that they belong to you — not fuel for someone else's platform, not hostages if you ever migrate.

The engine is the same one that runs real agents. This matters more than it first appears. letchat isn't a thin FAQ wrapper — it's the customer-facing surface of the same engine that powers rysh's terminal agents and channel humanoids, with the same design values: grounded answers, observable behavior, humans holding the last word. Your chatbot today can grow into email support and internal agents tomorrow, on one platform, instead of three more vendor subscriptions.

For agencies, that combination — white-label widget, custom domains, logic-in-code, data ownership — is a service line: deploy and manage chatbots for clients on infrastructure and terms you control, per client, without renting your margin to an enterprise suite.

Who this is for

If you've been quoted per-resolution pricing that made your eyes water, or you're three weekends into a DIY bot that still hallucinates your refund policy — you're the person we built this for.

Honest scope

The house style on these posts is a straight answer about maturity, so: letchat is early, and we're working with early customers closely rather than pretending to be a finished enterprise suite — that's a feature if you want influence over the roadmap, and a caveat if you need something battle-worn today. Under the hood it runs on Anthropic's Claude. And the ownership claims are the real thing: logic in code, your data yours, human takeover always available.

The chatbot on your website is the first impression your company's AI makes. It's worth owning.


Want a chatbot you own on your site — or want to offer them to your clients? We're onboarding design partners; early partners get direct input on the letchat roadmap.

Related: Humanoids: agents with a Slack handle and an email address · governance ai|human: the draft-and-confirm switch

Try Rysh

Every pane is a shell and an AI agent — install takes one command.

Get started free →