Spaces

Can you run two AI agents on one WhatsApp or Instagram account?

Yes. Spaces lets more than one AI agent work on the same connected account, and plenty of owners end up doing it without planning to: Claude in the browser for their own catch-ups, and an always-on agent such as OpenClaw or Hermes answering customers at night. It works well when each agent has a clear job and the right access, and it gets messy when two agents answer the same customer. This guide explains what each agent gets, how the account's limits are shared between them, a few ways to split the work, and the settings that keep them out of each other's way.

Why owners run two agents

There are a few common reasons, and they lead to different set-ups.

SituationA typical pair
You want replies at night but like to sit down with an agent yourself in the day.Claude or ChatGPT for you, OpenClaw or Hermes for after hours.
Two people in the business each prefer a different agent.One person on ChatGPT, the other on Claude, on the same accounts.
You want one agent to answer and another only to read and report.An always-on agent that replies, and Claude Code on a schedule that writes you a morning summary.
You are trying a new agent before switching.The old agent carries on, the new one reads only for a week.
A freelancer helps with one channel.Your agent on everything, theirs on one account only.

What each agent gets on its own

Every AI agent that signs in to Spaces, or uses a Spaces key, shows up separately in your dashboard under Your agent, in the list of agents with access. Each one opens to its own settings:

  • What it may do: read messages and accounts (always on), reply and send messages, post, comment and like, connect and disconnect accounts, run ads, and manage other agents.
  • Accounts it can use: every connected account, or only the ones you tick.
  • Daily share: how many messages and other actions this agent may use each day.
  • Pause, Resume and Remove: an off switch for this agent alone.
  • What it did lately: its last actions, so you can see which agent did what.

That means you can give an agent read-only access by leaving "Reply and send messages" unticked, keep it to one account, or pause it for the weekend, without touching the other agent.

The account limits are shared

This is the part people miss. The safe limits belong to the account, not the agent. Your WhatsApp number can send up to 400 messages a day and start 10 new chats a day once it has warmed up, at least 10 seconds apart, however many agents are using it. Two agents on one number do not double that. They share it.

A daily share is how you divide it. For example, on a WhatsApp number you might give the after-hours agent a share of 80 messages and leave the rest for the agent you use in the day. When an agent uses up its share, Spaces holds its next message and says when it can try again, while the other agent carries on. The account limits always apply on top, so no share can go above what the account can safely send. Instagram works the same way, within its own limits of up to 150 messages a day.

A reply to someone who wrote in recently is never held back by the account's daily limits, so customers still get answers. The shares matter most for new chats and busy days.

Ways to split the work

SplitHow it worksGood for
By timeOne agent answers in the day, the other in the evening and at night. Each has a rule in its notes about when to answer.Owners who want cover around the clock.
By roleOne agent replies; the other may only read and reports to you.A morning summary without a second voice in the chats.
By channelOne agent has the WhatsApp number, the other the Instagram account, using Accounts it can use.Businesses where each channel has a different job.
By personEach team member uses their own agent, and chats have an owner.Small teams sharing one set of accounts.

Splitting by role is the safest first step, because only one agent ever writes to customers. Splitting by time is the most common once you trust both agents.

Stop them answering the same chat

The real risk with two agents is a customer getting two answers, or two different answers, to one question. These habits prevent it:

  • Give each agent a clear rule in its notes about which chats are its job, such as "only answer between 6pm and 8am" or "only answer chats with no owner".
  • Use chat owners. Spaces can give each conversation an owner, and an agent can be told to leave chats that belong to someone else.
  • Hand chats to a person when you step in. A handed-over chat gets no replies, scheduled messages or routines from any agent until it is handed back.
  • Keep your facts in the Spaces business brief, which both agents read, so they never disagree about prices or hours.
  • Use different ways to talk to each agent, so a message meant for one never lands in a customer chat.

A worked example: day and night agents

Here is how a two-agent set-up might look for a small physiotherapy clinic. The owner uses Claude in the day, and Hermes runs on a small computer in the back office for evenings. Names and details are examples.

AgentWhat it may doAccountsDaily shareIts job
Claude (owner)Read, reply and sendWhatsApp and InstagramLeft at the account's full limitsCatch-ups at 9am, 1pm and 5pm with the owner.
HermesRead, reply and sendWhatsApp only60 messagesAnswers simple questions between 6pm and 8am, hands the rest to the owner.
WhoMessage
Patient, 9.40pmHi, do you have anything on Saturday morning? My knee is playing up again.
HermesHi Jo, sorry to hear that. I cannot see the diary tonight, but the team will check Saturday morning first thing and message you back. If the pain is severe or sudden, please call NHS 111.
Claude, to the owner at 9amOvernight, Hermes answered three chats. Jo is waiting to hear about Saturday morning for her knee. Hermes said the team would check first thing.
OwnerSaturday 10am is free. Reply to Jo and offer it.

Both agents read the same brief, so Hermes gave the clinic's standard line about urgent pain, and Claude picked up exactly where Hermes left off.

Notes for Hermes (evenings):
Only answer between 6pm and 8am. Outside those hours, do nothing.
Answer on your own: opening hours, prices from the brief, directions, parking.
Never book or move appointments. Say the team will confirm first thing.
Anything about pain, injury or treatment: give the brief's urgent line and hand the chat to a person.
Leave any chat that has an owner or has been handed to a person.

Who manages whom

One of the permissions is "Manage other agents". An agent with it can see your other agents, change what they may do and their daily share, pause or resume them, remove them and read what they did lately. It cannot change its own access or give away more than it has.

It is handy: you can ask Claude "pause Hermes until Monday" from your phone. Give it only to the agent you trust most, usually the one you talk to yourself, and never to an agent that runs unattended.

Checking who did what

When two agents share an account, a reply you did not expect raises the question "which one sent that?". Open the agent in Your agent and look at what it did lately, or ask the agent that manages the others to show you. The activity list shows the actions and the account, without the message text.

If one agent misbehaves, pause just that agent. A paused agent can read and send nothing until you resume it, and any scheduled messages or routines it set up wait too. The other agent carries on as normal.

Common mistakes with two agents

  • Both agents answering every chat, with no rule about whose job it is.
  • Keeping prices in each agent's own notes, so they drift apart.
  • Forgetting the limits are shared and wondering why the day agent is being held back.
  • Giving "Manage other agents" to the always-on agent.
  • Talking to an agent on the business number, so your instructions end up in a customer chat.
  • Adding a second agent to fix a problem with the first. Fix the first agent's brief and rules instead.

Summing up

Two AI agents can share one WhatsApp or Instagram account through Spaces, each with its own permissions, accounts, daily share and off switch. The account's safe limits are shared between them, so divide them with daily shares. Give each agent a clear job, keep the facts in the business brief they both read, use chat owners and hand-overs, and give "Manage other agents" only to the one you trust most.

Pricing

Plans from $19 a month

FAQ

Do two agents get twice the limits?

No. The limits belong to the account, so two agents share them. Use daily shares to decide how much each one may use.

Can I make one agent read-only?

Yes. In Your agent, leave "Reply and send messages" unticked for that agent. It can still read and summarise, but anything it tries to send is refused.

Can each agent work on different accounts?

Yes. Each agent has its own list of accounts it can use, so you can keep one to the WhatsApp number and the other to Instagram.

What happens if both agents reply to the same customer?

Both replies go out, as they would if two people answered. Prevent it with clear rules about whose job each chat is, chat owners and hand-overs.

Can one agent switch the other off?

Only if you give it the "Manage other agents" permission. Then it can pause, resume or change the other agent, but never its own access.

Do both agents need their own business notes?

Keep the shared facts in the Spaces business brief, which both read. Each agent's own notes should only say what is different about its job, such as its hours.

Is there a limit to how many agents I can connect?

Each agent that signs in or uses a key shows up in Your agent with its own settings. In practice, two is plenty for most small businesses, and each extra agent is one more thing to keep in step.

Can a team member use their own AI agent on our accounts?

Yes. Their agent signs in to Spaces and shows up in your list, where you choose what it may do and which accounts it may use. You can remove it at any time.

Related

More articles

AI agentsSigns your AI agent is sending too much, and how to rein it in11 Nov 2026 · 8 min readAI agentsWhat is an always-on AI agent?11 Nov 2026 · 10 min readAI agentsWhat is a draft-only reply?11 Nov 2026 · 8 min read