A modern support platform with AI at its core. An AI agent answers your customers, acts in your app, and brings in your team when it needs a person. Self-host it or use our cloud.

CM-1 Support exchange. Press a channel to send a message
AI agentWebsite

How long do Stripe payouts take?

Usually 2 to 3 business days after your client pays.

Source: Stripe payouts
YouIdle

Nothing needs you right now.

Go build your product.

Open-source AI agentOpen core, Apache-2.0Self-host or use our cloudYour own model key9 channels, one inboxHuman handover built inChats encrypted at rest
Three jobs, one AI agent
  1. Answers
  2. Acts
  3. Hands off
Chat widget on a sample website answering a question about payouts, with its sources listed

It learns your site, docs and PDFs, then answers from them.

Think of it as an AI chatbot trained on your docs. Paste a URL and it reads your pages, and you can check and edit everything it knows. Turn on sources to show them in the website chat.

Can you move our invoices to ops@studio-orchard.example?

Update billing email
Frommaya@studio-orchard.example
Toops@studio-orchard.example

Waiting for the customer to approve.

It does the work in your app, and asks first when it should.

This is what makes it an AI agent. Actions run on your own site, and you set which ones need approval. You can also send private details to a form on your site instead of the chat. Try the button above.

Shared inbox showing a chat handed from the AI to a person, with the whole conversation visible

When it needs a person, you get the whole chat.

Human handover is built in. Turn it on and the AI agent checks your hours and who is online, and you take over from a shared inbox on every channel.

Set up your AI agent in three steps

Step 1TeachAdd your website or docs.
Step 2TestAsk it what your customers ask.
Step 3LaunchPaste one snippet, or connect a channel.

Teach it once. Answer wherever customers write.

Add live chat to your website, then connect WhatsApp, Instagram, Messenger and more. Your customer support stays in one place.

On all 9 channelsAI answers, handover to a person, and one shared inbox for your team. See every channel
On your website tooLive chat, workflows, in-app actions, journeys and ratings.

Tickets that investigate themselves

When a chat needs more than an answer, it becomes a ticket. The AI tests its ideas against your logs and data, shows the evidence, and drafts the root cause. You choose how far it goes: investigate only, propose a fix for your approval, or resolve simple cases.

GrafanaSentryCloudWatchElasticsearchRead-only SQL

TKT-2041 Payout in transit past 3 days

idea 1 Payout is held at the bank

check get_payout: in_transit, day 4

result supported by payout record

idea 2 Bank details changed this week

check SELECT changed_at FROM bank_accounts

result ruled out, no change in 90 days

root cause drafted, waiting for your approval

Read it. Run it. Change it.

Every line of the AI agent, the inbox and the tickets is on GitHub under Apache-2.0. Teams leaving Intercom or Zendesk can run it themselves.

Your servers

Run it with Docker. Your data sits on your servers, and a local model keeps AI calls in house too.

Your model

OpenAI, Anthropic, Gemini, Mistral, or any OpenAI-compatible endpoint.

Your tools

Connect any MCP server, or call your own functions from the JavaScript API.

npm install -g chattermate-deploy && chattermate-deploy init my-chattermate && cd my-chattermate && chattermate-deploy start

Open core: the AI agent, inbox, channels and tickets are in the open-source repo. Our cloud adds the managed model, the CLI and MCP server, and the Shopify app. What is open

Software companies, online stores and service businesses

Each one starts from the questions that business gets most, and the tools it already uses.

Pay for what your AI answers

One AI reply is one message, and paid plans have unlimited team members. See full pricing

800
Free$050 messages a month
Starter$4.991,000 messages
Growth$9.992,500 messages
Pro$24.997,500 messages
Run it yourself$0Your servers, your ops

Paid plans have unlimited team members, and replies on your own key are not metered.

Fair questions

Is this customer support software really open source?

It is open core. The AI agent, inbox, channels and ticketing are open source under Apache-2.0. Our cloud adds a few hosted extras such as the managed model and the CLI.

Is it a chatbot or an AI agent?

Both. It works as an AI chatbot for customer service, answering from your content. It is also an AI agent, because it can run actions in your app, and either open tickets or hand chats to your team.

Does it replace live chat?

It includes live chat. The AI answers first, and your team can reply in the same conversation at any time from the shared inbox.

What happens when it doesn't know?

With handover on, it brings in your team with the full conversation. Outside your hours, or when nobody is online, it says the team will follow up and queues the chat.

Can I use my own AI model?

Yes, on paid plans and when you self-host. Add your key for OpenAI, Anthropic, Gemini, Mistral and others, or point it at any OpenAI-compatible endpoint.

Where does my data live?

On our cloud, or on your own servers if you self-host. Chats are encrypted at rest. The security page lists what is covered and what isn't certified.

Plug it in

Add your website and start teaching your AI agent.

Start free