CampaignStack vs Dux-Soup: the veteran desktop app against the platform

Dux-Soup has outlived most of the category it helped create. It has automated LinkedIn for more than ten years and claims 300,000 customers. In 2026 it went further than most vendors would, retiring the Chrome extension its name was built on and shipping a desktop app with a cloud option. Its default throttles are genuinely conservative: visits pace between 25 and 125 profiles an hour with a five-minute pause after every 20, connection requests default to 20 a day, its help center tells new accounts to start at 5 to 10, and it states it has had zero ban reports from users who stayed inside the recommended settings.

The July 2026 MCP server deserves its own mention, and it now comes with both plans. It lets Claude, ChatGPT or Gemini read campaign results, create and edit campaigns, manage prospects and draft replies, with permissions you choose when you connect it. Whatever else differs below, that is a real piece of engineering aimed at the right future.

STACKED local is a desktop app that runs the automation on your own computer, and Dux-Soup's help center states the cost plainly: for scheduled automation to run, the computer must stay switched on and not go to sleep. Campaigns run at the pace of your working hours on your machine, and a closed laptop lid ends the day's outreach. STACKED cloud exists to solve exactly this, at $99 a month per seat against $55 for local, and it starts by copying the LinkedIn session from your desktop app to Dux-Soup's servers.

CampaignStack runs every account in a persistent, encrypted browser profile on our servers, on the free plan and the paid one alike. Your machine is never part of the infrastructure, sends pace through business hours in the lead's timezone whether your laptop is open or not, and connecting an account takes a session cookie rather than a desktop install per seat.

Dux-Soup's safety documentation is among the more honest in the category: default limits matched to your LinkedIn account type, randomized daily activity, an automatic snooze at your daily cap or at LinkedIn's weekly invite limit, and a planner for working hours. But the daily caps remain settings. Its help center points you to the throttling tab to raise them, and its own caveat is that staying safe means staying inside the recommended settings. The defaults are a good start; past them, the safety system is you.

CampaignStack enforces its numbers. Every account gets per-action-type daily budgets capped below LinkedIn's practical limits, a warm-up ramp that grows caps with real usage, weekly connection caps, and an automatic risk pause when LinkedIn pushes back. A client cannot ask you to turn it up past safe, because there is no slider that goes there.

Dux-Soup executes the list you feed it. Building that list is manual by design: even its MCP guide is explicit that the AI cannot reach LinkedIn search, so someone runs the search, exports the results and hands them over. Once running, the AI campaigns write and send in one motion; no critic reads the draft, and no queue holds the doubtful ones for a human. The MCP lets you review a reply you asked it to draft before it is queued, which is a different thing from reviewing what each prospect in a running campaign is about to receive.

CampaignStack fills that gap with the deciding layer: signals update lead records, ICP scores gate who enters a campaign at all, every AI draft faces a critic before it ships, and the ones that keep failing wait in a review queue where approving releases the send. An agent driving our MCP server gets 100+ customer-facing tools that cover sourcing, scoring, drafting and review, so the part Dux-Soup leaves manual is the part an agent here can actually run.