Run cold email outreach with Meta Muse Code

Before you start

This page assumes Muse Code is already connected. If it is not, Connect WarmySender to Muse Code is three steps: create a key in Settings → API Keys, add one block to Muse Code's settings file, restart. Two things worth knowing before you read on:

• WarmySender is not a built-in Muse connector. Meta does not run a catalogue you can install us from, so there is nothing to click — the connection is one block you add to your own settings file on your own machine.
• Muse Code connects with a workspace key, not a sign-in screen. It has no browser sign-in step for MCP servers, so a key is the route we publish for it. A key reaches the same workspace, obeys the same limits, and can be revoked whenever you like.

One more thing we would rather say than imply: we tested the connection itself against Muse Code, and Muse Code really does reach the WarmySender server on startup. We have not run a WarmySender tool call from inside Muse Code ourselves — the prompts below are the plain-language shape of tools the WarmySender server exposes to every agent, not a transcript of a session we recorded.

What Muse Code can reach

Everything in this list is a tool the WarmySender server exposes. Nothing here is specific to Muse Code — any agent that speaks MCP gets exactly the same set.

• Cold email & prospects — Build, launch, pause, and resume cold-email campaigns; add, update, and bulk-import prospects; and manage your suppression list — all in plain language.
• Tracking domains — Use your own branded domain for open and click tracking. Add a tracking domain, check that it is set up correctly, list the ones you have, and remove any you no longer need.
• LinkedIn — campaigns, posts & recruiter — Run LinkedIn end to end: connection and message campaigns, publish and manage posts, and create, publish, or close recruiter jobs and read their applicants. Every action is paced within your account’s safe daily and weekly limits. Every action here is handed to the scheduler and paced; your agent cannot make it go faster.
• WhatsApp messages — Create and launch WhatsApp campaigns, enroll prospects, and check number health and results. New numbers ramp up gradually and stay within a safe daily, first-message and hourly budget, with a natural pause between messages. Every action here is handed to the scheduler and paced; your agent cannot make it go faster.
• Instagram DMs — Create and launch Instagram DM campaigns, enroll prospects, and check account health and results. New accounts ramp up gradually and stay within a safe daily and hourly action budget. Every action here is handed to the scheduler and paced; your agent cannot make it go faster.
• Warmup & mailboxes — Tune warmup settings, adjust many mailboxes at once, and check mailbox health and warmup ramp status before you scale up sending.
• Email verification — Verify a single address or submit a whole list, check how much of your allowance is left before you spend it, follow results as they come in, stop a run and get the unused credits back, and turn the usable results into a ready-to-use list — reported as confirmed mailboxes plus ones on mail servers that accept every address, so your agent knows what it actually has. All on the same safe pacing and allowance as the in-app verifier, so an agent can’t burst it.
• Leads database — Search 200M+ B2B leads by role, industry, and location, then save the ones you want into a prospect list or export them with full contact details. Searching is free; each new lead uses your monthly allowance or purchasable lead credits.
• Workspace & jobs — Confirm what an API key can access, set the postal address, opt-out style and opt-out wording used in your cold email footer, and track the status of any queued job or bulk operation.

Cold email, start to finish

The shortest useful path is four sentences typed one after another. You do not have to phrase them the way we have; naming the list, the steps and the timing is what makes a prompt land.

1. "Check my verification allowance, then verify this list of 800 addresses and tell me how many came back safe to send."
2. "Make a prospect list out of the ones that came back clean."
3. "Build a campaign called 'Agency founders Q4' for that list. Step one: a short intro email. Step two, three days later if they haven't replied: a one-line nudge. Weekdays only, send from my two warmed mailboxes."
4. "Start it, and tomorrow tell me what went out and who replied."

Step four is the one people expect to be refused, and it is not. Starting a campaign is allowed precisely because it does not send anything: it writes the campaign down and hands it to the scheduler, which then spreads the sends across your sending window, respects each mailbox's daily limit, keeps the whole domain under its ceiling, and skips anyone on your suppression list.

Useful follow-ups in the same session: "Why did that campaign send nothing yesterday?", "Pause it until Monday", "Which of my mailboxes look weak?", and "Use my saved template instead of writing a new email."

LinkedIn, Instagram and WhatsApp

Same conversation, different channel. Your agent can create, launch, pause, resume and enroll on each of them, publish LinkedIn posts, run recruiter jobs, and report on how each account is doing.

• LinkedIn — "Create a LinkedIn campaign on my main account: connect first, then message three days after they accept. Enroll these 60 profiles and start it."
• Instagram — "Build an Instagram DM campaign for this list and start it, then tell me tomorrow how many were delivered."
• WhatsApp — "Launch a WhatsApp campaign on my connected number for these 40 people and show me the results on Friday."
• All at once — "Build a campaign that emails them first, sends a LinkedIn connection request three days later if they haven't replied, and messages them four days after they accept."

One thing worth being blunt about: running four channels does not let you send more. Each channel keeps its own budget and none of them pool. Adding LinkedIn does not give your mailbox extra room, and adding Instagram does not lift LinkedIn's ceiling. What multichannel buys you is relevance and timing, not volume.

Each of these channels also needs its account connected first, and connecting always happens in the app — never through the agent.

Warmup and mailbox health

Warmup is the part people forget to ask about, and it is the part that decides whether any of the above lands. Ask for it directly: "Turn warmup on for my two new mailboxes and set a sensible target volume", "Apply the same warmup settings across all my mailboxes", or "How is reputation building on the mailbox I added last week?"

Your agent can set warmup up and read how it is going. It cannot shorten the ramp — a new mailbox starts slow on purpose and speeds up as it earns it, and there is nothing an agent can call that changes that.

Why a terminal agent suits this work

Outreach is repetitive in a way that rewards a terminal. The same four sentences run every Monday. The same "why did this send nothing?" question gets asked every morning. A terminal agent is already sitting in the folder where your lists live, it does not need a browser window waiting for you to click approve, and a key in a settings file is a better fit than a sign-in flow for a machine you leave running.

It also keeps the boring safety story intact. Whether you drive WarmySender from the app, from a terminal, or from Claude, ChatGPT, Cursor, Codex, OpenClaw, and Hermes Agent, the scheduler is the thing that decides what actually goes out, and it behaves identically in every case.

The line your agent cannot cross

Your agent never sends a message, DM or invite itself, and it can never raise a limit. Creating or launching a campaign only writes it down and hands it to WarmySender's scheduler, which paces every email and every LinkedIn, Instagram and WhatsApp action inside safe daily caps and the gradual ramp — the same protections the app uses, whether a human or an agent pressed go. Ask it to enroll a huge list at once and you get a queue that drains at your normal safe pace, not a burst. Email verification shares the same allowance and pacing. And connecting or disconnecting your accounts stays in the app — your mailbox and social logins never pass through an agent or a chat log. Account safety always comes first.

If your agent misunderstands you, nothing goes out in a burst, pausing is one sentence and is never blocked, and everything it built is an ordinary campaign you can open, edit or delete in the app. For a cautious first week, create a read-only key in Settings → API Keys and let it look before it touches.

Frequently asked questions

Can Muse Code send my cold emails for me?

No agent sends anything, and that is deliberate. Muse Code can build the campaign, add the people and start it — and starting it simply hands it to WarmySender's scheduler, which paces every send inside safe daily caps and a gradual ramp. It also cannot raise a limit and cannot connect or disconnect a mailbox or a social account.

Is WarmySender a built-in Muse Code connector?

No. Meta does not run a third-party catalogue you could install WarmySender from, so there is nothing to click. The connection is one block you add to Muse Code's own settings file on your own machine, using a key you create in WarmySender. It takes about a minute and is covered step by step in our Muse Code setup guide.

Have you actually run WarmySender from Muse Code?

We tested the connection, and it connects: Muse Code reads the settings block, reaches the WarmySender server on startup and reports back on the key. We have not run a WarmySender tool call from inside Muse Code ourselves, so we do not claim one. The tools themselves are the same ones every other connected agent uses every day.

Can it run LinkedIn, Instagram and WhatsApp too, or just email?

All of them. It can create, launch, pause, resume and enroll on cold email, LinkedIn, Instagram and WhatsApp, publish LinkedIn posts and run recruiter jobs, and report on each. Each channel needs its account connected in the app first, and each keeps its own limits — adding channels adds touchpoints, never headroom.

I asked it to start a campaign and nothing sent. Is it broken?

Usually not. Ask it directly — "why did that campaign send nothing?" — and it will tell you in plain words: the sending window had not opened yet, the daily limit was already used, the mailbox is still early in its warmup ramp, people were skipped because they are on your suppression list, or a campaign start date is still in the future. A first send being paced rather than immediate is normal, not a fault.

Does running outreach through Muse Code cost extra?

No. Connecting an AI agent is included with any active WarmySender plan — there is no separate add-on to buy. Your agent draws on the same allowances you already have, and cannot spend them any faster than you could in the app.

Related reading

• Connect WarmySender to Muse Code — the three-step setup, and the two things that usually break it
• Can Meta Muse run my outreach? — Muse the consumer assistant, as opposed to Muse Code the terminal agent
• What your AI agent can and cannot do with cold email — the full capability matrix
• Which AI agents can run outreach — the per-app, per-plan comparison
• Connect an AI agent — one-paste setup for every supported agent
• The WarmySender MCP server — the full tool reference