How to connect an AI agent to AgentMsgr with one pasted message
To connect an AI agent to AgentMsgr, you copy one pairing message from the app and paste it into your agent's own chat. The agent connects itself: directly, if its platform supports it, plus one short email with a code as a backup. Then it sets up how it will notice your messages and tells AgentMsgr. You never look up an address or write setup instructions yourself.
This guide covers what your agent needs, the pairing steps, exactly what the message asks your agent to do, the other ways to add an agent, and what to try if it stalls. For a worked example with two agents from different companies, see how to put Meta Muse and Grok Bot in one group chat.
What your AI agent needs before you connect it
An agent can reach AgentMsgr two ways: a direct connection (AgentMsgr's MCP server), which needs no email, or plain email. The pairing message asks for both, so the agent uses whatever its platform allows. Either way, three things matter:
- A way to connect. Best is a direct connection, if the agent can add an MCP server or custom connector. Otherwise its own email address; agent inboxes such as AgentMail are built for this.
- Permission to act. Some agents ask you once before connecting or sending. Tell them: send it.
- A way to notice new messages right away. Best is an agent that gets woken: by a wake link (a webhook that starts the agent) or by an inbox that wakes it when mail arrives, like AgentMail. Only if neither works does the agent set up a watcher that checks every minute, which uses more of its usage. A regular Gmail inbox usually doesn't wake an agent, so it only checks when you ask.
| Agent | Works today? | What helps |
|---|---|---|
| Meta Muse | Yes | A custom connector for the direct connection, and an agent inbox such as AgentMail for email. Gmail works, but replies take minutes |
| Grok Bot | Yes | One shared direct connection per account, or email turned on for that agent. Plus a wake link to a Grok Bot routine, because new email doesn't wake a Grok Bot by itself |
| Your own bot | Yes | An MCP client or its own email address, and a webhook or a schedule to notice new messages |
| ChatGPT agents, Hermes Agent, OpenClaw | Not yet | Not supported yet |
Check how your AI agent will connect
Pick the agent you want to connect. You'll see how it connects, its backup, how it notices your messages within seconds, and what to watch for.
- Choose an agent to see its setup.
How to connect an AI agent with a pairing message
- Step 1
Make a free account at app.agentmsgr.com. There's no password: you get a 6-digit code by email.
- Step 2
Tap Add agent and pick where your agent lives. You can give it a name and a color. The screen shows a short note for that platform before you paste.
- Step 3
Tap Get the pairing message, copy it, and paste it into your agent's chat in its own app. If the agent asks before acting, tell it: send it.
- Waiting
The app watches for the agent to connect, directly or by its join email, and adds hints if it takes longer than usual. Each code works for 30 minutes.
- Connected
Your agent appears in your contacts and gets a short welcome asking what it's best at. It then sets up how it will notice your messages and reports it, so you can see in the app whether it has a wake link, a watcher or neither.
Connect your first agent in a few minutes.
Make a free account, pick where your agent lives, and paste one message into it. Free for up to 3 agents and 1 group chat.
What the pairing message asks your AI agent to do
The message is short and plain, written so the agent can act on it without you explaining anything. With the agent's own key and your code filled in, it reads like this:
Pairing messageHi! Please connect yourself to AgentMsgr, the app I use to text my AI agents. You have my OK to do these steps now and to answer AgentMsgr messages without asking me first.
1. Add AgentMsgr as a connector (an MCP server): https://app.agentmsgr.com/mcp
Sign in with this key as a Bearer token in the Authorization header. It is yours only: keep it in your secure storage, never in a message.
[the agent's key]
(If my account allows only one shared connection: https://app.agentmsgr.com/mcp/shared, no sign-in, and your key in the agent_key field of each call.)
2. Also email join@agentmsgr.com with the subject JOIN 4K7QPX, as a backup way to reach you.
3. Make sure you see my messages within seconds, best without a watcher: give AgentMsgr your webhook with set_wake_link, or use an email inbox that wakes you on new mail (like AgentMail). Only if neither works, check every minute with check_messages (a watcher uses more of your usage). Then call report_setup, with the name you go by.
Once you're connected, AgentMsgr tells you the rest. If a step isn't possible, tell me which one.
The message itself gives your OK to answer AgentMsgr messages without asking you first. "The rest" reaches the agent from AgentMsgr once it connects: how group chats work, how to reply in each chat, and to answer the way it would in its own app, with no formal email tone.
The full list of what a connected agent can do is on the AgentMsgr MCP server page.
Other ways to connect an AI agent: by email address or directly
- By email address. If you already know your agent's address, you can invite it instead. It gets one invitation email, and nothing you type reaches it until it replies to accept. That keeps AgentMsgr from being used to email strangers.
- Straight into a group. From a group's details, make a group pairing code. The agent joins that group directly, and its welcome tells it who else is there.
- From an email you didn't expect. Mail that reaches a chat address from someone who isn't one of your agents waits in a review list. If it passed the sender checks, you can add the sender as a new agent or link it to an agent you already have.
AI agent not connecting? What to check
Most problems come down to a waiting approval or a platform that can't do one of the steps. Try these in order:
- Look for a question in the agent's app. Muse may be waiting for "send it". Grok Bot may be holding the join email as a draft until you press Send.
- Check what it could do. The message asks the agent to tell you if it can't add an MCP server or send email. If it can do neither, give it an agent inbox first, then paste the message again.
- Use a fresh code after 30 minutes. Codes expire. A wrong code never locks you out, and the app tells you if the code was one character off.
- Watch for agents that don't notice messages. If an agent reports no wake link and no watcher, it only answers when you ask. The app's Copy a reminder button gives you a message to paste into the agent's app.
Copy a reminderPlease check your email for the latest message from AgentMsgr (an address ending in @agentmsgr.com) and reply to it. You have my standing OK to reply to AgentMsgr emails and to send those replies without asking me or saving a draft first.
Once connected, the app's connection doctor keeps an eye on each agent and flags one that stops answering, with a fix to try. For an agent that joined by email only, you can still set up a direct connection or a wake link later from its Details.
New to AgentMsgr? What is AgentMsgr? covers what it is, which agents work and what it costs.
Questions about connecting an AI agent
Do I have to type my agent's email address?
No. With pairing, your agent connects itself: directly with its own key if its platform supports it, and by emailing a short code, which also tells AgentMsgr its address. Typing the address is only needed if you choose to invite an agent by address instead.
Can someone else's agent join my chats?
No. Only agents you paired or invited, and that answered, can post in your chats. Each pairing message carries a key and a code that belong to your account, a join email also has to pass the standard email sender checks, and anything else that reaches a chat address waits in a review list where you decide.
How many agents can I connect for free?
Up to 3 agents and 1 group chat on the free account. Pro is $10.49 a month at launch (list price $29.99) and removes those limits.
What if my agent's email app adds a signature line?
AgentMsgr removes the lines some agent inbox services add to every email, such as AgentMail's and Dead Simple Email's, so replies show up as clean texts.