Campaign Swarm researches your market, plans the campaign, produces the assets and runs the ads — then stops at your desk for the decisions that spend money. A swarm of specialists, one approval gate, and nothing ships that you did not okay.
researching the weekend-improver audience…live swarm
Watch how Campaign Swarm plans the strategy, generates the creative, and watches every campaign around the clock.
Campaign Swarm
A Swarm of AI Agents
That plans, creates, and runs your campaigns for you.
One sentence. One click. Full campaign.
campaignswarm.com
2026
The Problem
Your team works 9 to 5
Your campaigns run 24/7. Your team doesn't.
Manual Process
StrategyWeeks
CreativeDays
BuildHours
The 6PM Gap
6:00 PM
Team clocks out
Meta
TikTok
Spending your budget while you sleep
The Swarm
Three products. One swarm.
Forty seven plus fourteen agents working together.
Strategy Lab
47 agents
Research audience, debate segments, build full strategy grounded in your data
Minutesinstead of weeks
Creative Studio
14 agents
Turn research into images and copy. Launch to Meta and TikTok with one click.
One Clickready to launch
Mission Control
Always on
Watches every campaign, every 30 minutes. Catches budget blowouts before they happen.
30 mindetection, not Monday
What This Means
More output. Same team.
The swarm multiplies what your agency can do.
More clients
Take on more clients without adding headcount.
Win pitches
Win pitches you used to hand-wave because you didn't have the bandwidth.
Growing margins
Your margins grow while your workload stays the same.
Getting sharper
Every campaign makes the next one sharper. The swarm learns from what actually worked.
Always in control
You're the human in the loop. Every AI decision is transparent, reviewable, and yours to approve or override.
Plan
/
Make
/
Run
That's what the swarm does.
The only thing you have to do is decide what it's allowed to.
The rest happens automatically.
Campaign Swarm
AI Campaign Intelligence Platform
campaignswarm.com
2026
Campaign Swarm
campaignswarm.com
Click to start presentation
0 / 6
The bind
You are the strategy. Also the account manager. Also the person fixing a client's caption at 11pm.
There is a planner, a scheduler, a design tool, an ad dashboard and a spreadsheet. None of them talk to each other, so the thinking lives in your head — and the moment a client asks for a change, the work restarts from scratch. Campaign Swarm holds the whole loop in one place and keeps it moving without you.
6+
tools stitched per client
4 hrs
a week lost to posting
0
of it remembered next month
How it works
01 / 05
It reads your market before it writes anything.
02 / 05
It writes a plan you can read in one sitting.
03 / 05
It builds the assets in each platform's own shape.
04 / 05
You hold the gate on anything that spends.
05 / 05
It ships, watches, and folds the result back in.
Four research agents fan out at once: one on your product and category, one on the objections buyers actually raise, one on what your competitors are saying, and one on the numbers you already have. They do not invent a market — every claim is tied to a source the plan was given, whether that is your site, your analytics or the notes you typed into the brief. The synthesis agent reconciles the four into a single brief, and the disagreements it could not settle are shown to you instead of averaged away.
Grounded in your pages, posts and first-party data
Sources cited per claim, not summarised away
Open questions surfaced, never silently resolved
The brief becomes an actual plan: the audience it is going after, the offer it will lead with, the angle it believes will land, the channels it will use and the budget it wants. It is one screen, not a deck. Because the plan is the unit of work, everything downstream — every caption, every image, every scheduled post and every ad — points back to the plan that produced it. When a post underperforms three weeks later, you can trace the assumption that led there and change it at the source.
Audience, offer and message decided together
A budget recommendation with its reasoning attached
Every asset traceable to the plan it came from
Copy, hooks and creative are produced per format — a 15-second reel script, a four-slide carousel, a one-line Threads post, a comment reply that sounds like a person. Three directions are generated from the same research, so you are choosing between real strategic options rather than variations in tone. Nothing starts from a blank prompt: the segment, the objection and the angle are already in the brief, which is why the output tends to sound like your business instead of like a language model.
Three directions per campaign, not one guess
Format-aware copy for feed, story, thread and reply
Editable at the sentence level, or regenerate a direction
Campaigns are created paused. Every post waits in a queue with its caption, its image and the reason it exists. Budget moves, pauses and creative refreshes arrive as requests with a confidence score, the evidence behind them and the guardrail that triggered the review. One tap approves; one note rejects, and that note is read back into the next plan rather than filed away. You can set the policy once — what may happen alone, what always needs you — and the swarm never steps outside it.
Campaigns born paused, never live by default
Rejections teach the next plan, not just this one
A policy you set, enforced on every action
Approved work publishes on a short cadence and the monitor starts reading the result the same day: reach, saves, clicks, cost per result, and the comments nobody has answered yet. Those numbers become the next plan's starting assumptions, so the account compounds instead of merely staying busy. You get a weekly digest that says what changed, what it thinks you should do next and why — in the same plain language as the rest of the product, with the audit trail underneath if you want to check its work.
Publishing and monitoring on one loop
Learnings applied to the next plan automatically
A weekly digest with the reasoning, not just the chart
Four research agents fan out at once: one on your product and category, one on the objections buyers actually raise, one on what your competitors are saying, and one on the numbers you already have. They do not invent a market — every claim is tied to a source the plan was given, whether that is your site, your analytics or the notes you typed into the brief. The synthesis agent reconciles the four into a single brief, and the disagreements it could not settle are shown to you instead of averaged away.
It writes a plan you can read in one sitting.
The brief becomes an actual plan: the audience it is going after, the offer it will lead with, the angle it believes will land, the channels it will use and the budget it wants. It is one screen, not a deck. Because the plan is the unit of work, everything downstream — every caption, every image, every scheduled post and every ad — points back to the plan that produced it. When a post underperforms three weeks later, you can trace the assumption that led there and change it at the source.
It builds the assets in each platform's own shape.
Copy, hooks and creative are produced per format — a 15-second reel script, a four-slide carousel, a one-line Threads post, a comment reply that sounds like a person. Three directions are generated from the same research, so you are choosing between real strategic options rather than variations in tone. Nothing starts from a blank prompt: the segment, the objection and the angle are already in the brief, which is why the output tends to sound like your business instead of like a language model.
You hold the gate on anything that spends.
Campaigns are created paused. Every post waits in a queue with its caption, its image and the reason it exists. Budget moves, pauses and creative refreshes arrive as requests with a confidence score, the evidence behind them and the guardrail that triggered the review. One tap approves; one note rejects, and that note is read back into the next plan rather than filed away. You can set the policy once — what may happen alone, what always needs you — and the swarm never steps outside it.
It ships, watches, and folds the result back in.
Approved work publishes on a short cadence and the monitor starts reading the result the same day: reach, saves, clicks, cost per result, and the comments nobody has answered yet. Those numbers become the next plan's starting assumptions, so the account compounds instead of merely staying busy. You get a weekly digest that says what changed, what it thinks you should do next and why — in the same plain language as the rest of the product, with the audit trail underneath if you want to check its work.
Who it is for
Built for the agency doing the work.
Not another per-seat tool your clients never see. A private, white-labelled deployment for the agency that owns the relationship, the roster and the standards.
Agency principals
You own the roster, the margin and the standard.
You are running five to thirty client brands and the platform is part of what you sell. Your logo, your domain and your client review portal — on a deployment of your own, priced per agency rather than per seat. One place to plan, produce, approve and prove the work.
Client service leads
The roster is yours; the context switching is not.
Every client gets a plan, a calendar, an approvals queue and an audit trail that explains why a decision was made. When someone is away, the reasoning is still there to read. You stop stitching five dashboards and three tools into one story for the client.
Studio leads
More output at the standard you already set.
The planner writes the argument down, the studio produces the variants, and the swarm handles scheduling and optimisation. You review the work instead of making every first draft. AI credits are pooled across the deployment, and top-ups never expire.
The rest of the workspace
One workspace. Research to reply.
Not a scheduler with AI bolted on. The strategy, the creative, the ads and the inbox all read from the same plan, which is why the fourth thing you write sounds like the first.
Creative Studio
Brief to finished campaign in one project.
Research, three strategic options, generated assets and a launch you control. The studio keeps the plan beside the canvas, so the hook you are editing still shows the objection it was written to answer. Fourteen specialists across five stages — copy, art direction, format adaptation, brand checks and compliance — all working from the same brief instead of passing a document around.
Per-format output: feed, story, thread, reply
Brand palette, logo and tone applied automatically
Predictive CTR / CVR bands before you publish
Mission Control
Every live campaign, and the reasoning behind it.
Pacing, spend, return, fatigue and the decision the monitor took — on one screen, with the reasoning written down. When a number moves, you can open the decision that responded to it and read the evidence, the confidence and the guardrail that applied. No black box, and no dashboard that shows a metric without saying what it means.
Blended return and cost per result, computed
Creative fatigue scored per asset
Every autonomous move linked to its evidence
Approvals
The human gate, as a screen rather than a promise.
What the swarm wants to do, why it wants to do it, and what it expects to happen. Approve in one tap, edit the instruction, or reject with a note that the next plan reads back. Nothing that spends money bypasses this queue, and the queue keeps the full history — who decided, when, and what happened afterwards.
Budget, pause and refresh requests in one inbox
Policy per action class, not a global toggle
Exportable audit trail for every decision
Organic & inbox
Posting and replying in the place you plan.
Instagram, Facebook and Threads in one calendar, with the comments that need a human kept beside them. A reply can be drafted from the same research that produced the post, so the voice stays consistent when someone asks a real question under it. Nothing is auto-sent; replies wait for the same gate as everything else.
One queue across every connected page
Comment drafts grounded in the original post
Unanswered threads surfaced as a weekly action
Content engine
Planner, studio, calendar and publish queue.
Approve once and the queue handles scheduling, retries and the honest reason when something fails — a token, a permission, a platform limit. The calendar is the same object the planner wrote, so moving a post does not fork the plan. Failed publishes come back with a classification, not a stack trace, and the fix is usually one click away.
Scheduling with per-account guardrails
Failure reasons in plain language
Retry without losing the asset or the slot
MCP-native
One key installs the workspace into your agent.
The same workspace is an MCP server: twenty-five tools that read accounts, posts, comments, ads, plans and digests, and write drafts, schedules and approvals. Claude Code, Codex, Cursor and VS Code can all drive it with a workspace key you can revoke. Reads stay read-only; writes still land in the approvals inbox, because the gate does not disappear when the interface does.
Read tools for every object in the workspace
Writes create drafts or approval items
Same key, same scopes, same audit trail
12
stages in one plan, argued in the open
3
channels publishing today, with ads in the same loop
25
MCP tools for Claude Code, Codex and Cursor
0
budget moves that happen without your approval
Channels & data
Where you already post. What you already know.
Connect your pages and your analytics once. The planner grounds its thinking in your real audience, and the queue publishes to the accounts you actually own.
Publishing today
Your social pages
InstagramFeed, carousel and Reels
FacebookPage posts and scheduling
ThreadsText posts and replies
Scheduled posts, replies to comments, and a calendar you can move things around in.
Ads today
Meta ads, created paused
Facebook & Instagram placements
Campaign, ad set, creative and ad built for you
Nothing spends until you press activate
Budget moves above your cap need approval
Spend is read back daily, so the next decision is based on how the last one actually performed.
Grounding
Your first-party data
PostHogProduct events and behaviour
HubSpotContacts, companies, deals
MixpanelProfiles and funnels
Read-only scopes. Attached per plan, never workspace-wide, so one brand's data cannot shape another's strategy.
On the roadmap — shown as roadmap, not as shipped
TikTok is next. LinkedIn, Reddit and Google Ads follow.
Publishing is scheduled; ads are created paused. Every budget change, pause or creative refresh arrives with the reasoning, the confidence and the evidence behind it. You approve, change or ignore — and the run continues on your terms.
Set the policy once: what the swarm may do alone, and what always needs a human. It never steps outside it, and everything it did is in one exportable history.
Campaigns are created paused, never live by default
Budget, pause and creative changes queue for a human
Rejections carry a note the next plan actually reads
Needs your approval
Shift budget to “Weekend Cram”
87% confident
Reasoning
Ad set B is holding a 2.1× return at a 22% lower cost per result over 7 days.
Evidence
1,284 clicks · 41 results · fatigue score 18/100
Guardrail
Budget shifts above 20% always need a human
A preview of the real approvals inbox. Every decision is kept.
MCP · for the people who live in a terminal
Your marketing, callable from Claude or Codex.
The same workspace is exposed as an MCP server, so you can ask your coding agent to check yesterday's reach, draft next week's posts or move a schedule — without opening a dashboard.
Reads are read-only; writes become drafts or approvals
Reach is up 18% week over week. The Tuesday reel did 12.4k with a 3.1% engagement rate — the best of the last 30 days.
> draft three follow-ups for next week
Drafted 3 posts, scheduled as drafts. Two need an image; one is ready for review in the approvals inbox.
list_accountsget_performance_digestlist_contentget_top_postslist_commentscreate_draftschedule_postreply_to_commentapply_performance_learningsupdate_schedule+15 more
Deployment pricing
Your agency. Your platform, not another seat.
Every deployment is private: your domain, your database, your storage and your encryption key, under your logo and your client review portal. Priced per agency, not per seat or a share of ad spend. Request the price sheet and we will confirm the tier for your roster.
Monthly or annual · AUD, ex GST · per deployment, never per seat
Every price is confirmed on a current price sheet, in AUD ex GST. The deployment includes approvals, audit trail, client review links, scheduling, inbox, analytics, MCP and Australian hosting; AI credits beyond the monthly allowance are prepaid and never expire. Annual is 10 months for 12. Ad spend is billed by Meta to the client's own ad account — never through us.
Questions
The things people ask before signing up.
You see it before it costs you anything. Posts are scheduled and ads are created paused, so a bad idea lands in your review queue rather than in public or in your ad account. Every autonomous decision also carries its reasoning, so you can correct the thinking and not just the output.
You do. Your posts, images, plans and performance data belong to your workspace. Connected analytics credentials are encrypted and used read-only; disconnecting a source removes its access immediately.
No. Campaign Swarm creates campaigns paused. You activate them. After that, budget moves within the cap you set can happen automatically, and anything above it becomes an approval. You can change the cap or pause everything from one screen.
Instagram, Facebook Pages and Threads today, plus Meta ads for Facebook and Instagram placements. TikTok is next; LinkedIn, Reddit and Google Ads are on the roadmap and are shown as such — we do not pretend they are live.
No. You answer questions about your business, your offer and your audience, and the planner does the rest. If you prefer to steer it in prose, the assistant takes instructions in plain language — and if you live in a terminal, the MCP server lets you run the same workspace from Claude Code or Codex.
One monthly fee per agency deployment, not per seat or a share of ad spend. Tiers are sized by client brands, from three to fifty, with a monthly AI credit allowance and prepaid top-ups that never expire. We send the current price sheet by email and confirm the tier for your roster. Nothing is charged through this site.
Month to month, or annual at 10 months for 12. Cancel with 30 days' notice. You can export your plans, posts, assets, analytics and decision log, with a restorable database dump, and deletion follows after 30 days.
We build your first client campaign with you in a 90-minute session. Ten founding deployments, white-label from day one, with your own database, storage and domain.