The AI support desk that
never makes things up
momo answers what it knows from your content, takes real actions like checking an order status, and hands everything else to a clean inbox your team will actually enjoy.
MIT
Open source license
<30KB
Widget script size
~1 min
To first answer
Why momo
Support bots make things up. momo is built around not doing that.
No resolution tax
Flat pricing, unlimited seats. Bring your own model key if you like — no per-resolution fees, no credit system.
It says "I'm not sure"
Every answer carries sources and a confidence level. Below the threshold, it hands off to a human instead of inventing a refund policy.
It does things, not just FAQs
"Where's my order?" gets answered live from your shop, not a canned response.
How it works
The same loop, every time: answer confidently or hand off honestly.
- 1
Paste your site or upload your docs
momo crawls your site, PDFs, or Q&A pairs and starts answering from your real content within about a minute.
- 2
The accuracy gate checks every draft
Before an answer ever reaches a visitor, it's checked against the sources it's supposedly drawn from.
- 3
Confident answers ship, the rest become tickets
Not confident enough? It says so, and the conversation lands in a clean inbox for a human — with a one-click Teach button that turns the reply into approved knowledge.
Every conversation
What happens when someone asks momo something
The same accuracy gate runs every time, whether the answer ships or a human takes over.
0-2s
momo drafts an answer from your sources
Pulled from your crawled site, docs, or Q&A pairs, never from general knowledge.
Gate
The accuracy gate checks it against those sources
If the draft is not backed by what it found, it does not ship.
Confident
A sourced answer reaches the visitor
With links back to the page it came from.
Not confident
It says so and hands off
The conversation becomes a ticket in the inbox instead of a guess.
Everything in the open core
Self-hosted, MIT-licensed, and honest: real services, no artificial limits.
- Website crawl, file upload, Q&A pairs, text snippets
- Answers with sources and an accuracy gate
- Embeddable widget, one script tag, under 30KB
- Shopify order-status action
- Inbox with live takeover and simple tickets
- Teach button on every human reply
- Unlimited team members, no seat cap
- REST API and webhooks
Open core, self-hosted
- MIT licensed, full source access
- Bring your own AI model key
- Unlimited team members
- Runs on your own infrastructure
Cloud, early access
- Same open-core loop, managed for you
- AI usage included in the price
- EU hosting
- Priority support
Why teams switch from Chatbase
A side-by-side look at what you pay and what you get.
Based on Chatbase's public pricing and product pages, momobot's own research (2026-09-21).
Open by default
No lock-in, no black box.
MIT licensed
Fork it, self-host it, read every line. No paid tier required to see the code.
Self-hosted or cloud
Run it on your own server with your own AI key, or use the managed cloud, same code either way.
Your data stays yours
Self-hosted, nothing leaves your infrastructure. On cloud, EU hosting and no training on your content.
Why we built this
Support bots that guess are worse than no bot at all
Most AI support tools are optimized to always answer, because a confident-sounding reply looks good in a demo. That is how you end up with a bot inventing a refund policy that does not exist.
momo is built around the opposite instinct. Every answer is checked against the sources it is supposedly drawn from before it ever reaches a visitor. If the confidence gate does not clear, momo says so and hands the conversation to a human instead of guessing.
We are building it in the open, MIT licensed, because a support desk that touches customer trust should not be a black box you have to take on faith.
Unlimited
seats, on every plan
No per-agent fee, no per-resolution credit system. Add your whole team without doing the math first.
Frequently asked questions
Is momo open source?
Yes. The core loop — the widget, answers with sources, the accuracy gate, the inbox with takeover, simple tickets and Teach — is free and self-hostable forever. The commercial cloud adds things that cost us money to run or connect to other tools, like AI keys included and JS-rendering crawling.
How is momo different from Chatbase?
Every answer carries sources and a confidence level. Below the threshold, momo says "I am not sure" and hands off to a human instead of inventing an answer. Pricing is flat with unlimited seats, not a per-resolution fee.
Do I need my own AI model key?
On the open core, yes — bring your own key for any provider supported by the Vercel AI SDK. The commercial cloud includes AI usage in the price, so no key is needed there.
What happens when the AI is not confident in an answer?
It says so in plain text — "passed to the team" — instead of guessing, and the conversation becomes a ticket in the inbox. A human replies, and one click (Teach) turns that reply into approved knowledge so the bot improves.
Build a support desk that doesn't lie to your customers
Get in touch — early access is open.