Announcement
Replaid Now Connects Your AI Agent to Your Customers
Bring the agent you already use to your customer conversations. Replaid connects the channels and lets you control what your agent can do.
On this page
Replaid connects your website, social, and messaging channels to the AI agent you already use. Your agent can read a conversation, prepare a reply, and send it through Replaid when your permissions allow it.
That is a change from the earlier product. Replaid started as an AI inbox assistant, with the AI presented as part of Replaid itself. The current product gives your own agent access to customer conversations.
Your agent brings the knowledge. Replaid provides the connection and action controls.
What changes for you
If you already use an agent with your business instructions, you can use that agent to help with customer messages. You do not need to treat Replaid as a second agent with a separate set of answers to maintain.
You choose the agent and give it the information it needs. Replaid connects supported channels, provides conversation context, and checks requests to act on those conversations against your policies.
The inbox remains a place to inspect conversations and review replies. The main purpose is to make those conversations available to your agent, wherever you run it.
A customer question, from message to reply
Here is an illustrative example, not a customer case study.
A visitor asks through your Replaid widget: “Do you offer Saturday appointments?” Your agent reads the conversation through Replaid and uses the opening hours you have already given it.
Start by asking it to save a draft. Review the answer in Replaid. If the agent does not have your opening hours, it should ask for them instead of inventing an answer.
When you are ready, you can permit sending and ask the agent to send an approved response through the same channel. Reading, drafting, and sending are separate steps. Connecting the agent does not give it unrestricted permission to reply.
The same approach can support other tasks: tagging a conversation, recording a lead qualification decision, or escalating a case for human attention. The agent must have the required permissions, and the action must be available for that conversation.
What connects today
The current channels are:
- WhatsApp: customer messages and text replies through your WhatsApp Business number.
- Replaid widget: customer conversations from your website.
- Instagram: a Business or Creator account, with text DMs and text comments.
- Messenger: messages to a Facebook Page.
- Telegram: private messages to your bot.
Each team can connect one Instagram account, one Facebook Page, one Telegram bot, and one Replaid widget. These four channels support text messages and replies; attachments are not supported on them. TikTok is not available yet.
Your agent connects through MCP or the REST API. The connection guide has setup instructions for supported clients and the current connection limits.
You decide when the agent runs
An MCP connection gives your agent tools. It does not start the agent when a message arrives, and it does not turn on automatic replies.
You can ask your agent to work on a conversation yourself. For unattended processing, you need an external schedule or a webhook receiver that starts your agent. Replaid provides the conversation access and permitted actions; your agent’s runtime handles when it runs.
Begin with a narrow task and draft review. Broader permissions can come later, after you have checked the results on your own messages.
Start with one channel and one draft
- Create your Replaid account, verify your email, and connect one channel when you are ready.
- Connect the agent you use and give it the relevant business knowledge.
- Ask it to read a conversation and save a draft without sending it.
- Review the draft, adjust its instructions, and test a permitted send when you are ready.
Create your account, or follow the step-by-step connection guide. Account creation is free; the pricing page explains paid usage and optional offers.
What happens to the older articles
The original launch announcement remains available as product history. It has a notice explaining that it describes the earlier version.
The practical guides now reflect the current product. Their original publication dates remain visible, alongside the date of the update. For a concrete next step, read three customer message workflows for your agent.