You describe the creator you want in plain sentences. The agent searches your niche, looks at each profile it finds, grades it and says why, then reaches out to the ones that clear your bar: one like, one comment, one follow, one message, spaced out and written in their language.
No scrolling profiles for days, no spreadsheet of handles, no copy-pasted DM. You describe the fit once; the agent finds, grades and approaches.
You describe the creator you want in a few plain sentences: who is a good fit for the brand, and who is not. That description is what every account gets graded against, so changing the shortlist later means editing a paragraph.
The agent searches the platform for the words and hashtags of your niche, and can work through the followers of competitor or aspirational accounts you name. Your own account is always left out.
For each creator it finds, the agent opens the profile and reads the actual page, the bio, the grid, the ratio of followers to following. Back comes a score out of ten, a letter grade, and one or two sentences explaining it. Fit is 40 percent of that, whether the account is real and active another 30, how likely they are to engage the last 30.
A creator who clears your score bar gets at most one of each, and never all four in an afternoon: around 36 hours sits between two actions on the same person. Replies land in one inbox, and the agent can answer them in your voice.
Who is a good creator for you and who is not, in two to four concrete sentences. That is the whole screening rule. There is no scoring model to configure and no saved segment to maintain, so tuning who gets approached is rewriting a paragraph.
Every account it finds is stored with a score, a letter grade, how sure the model was, and one or two sentences explaining the grade. Nobody is contacted unless their score clears the bar you set, seven out of ten by default, and the ones it passed on stay on the list with their score, so you can see who it said no to.
The grading is done from a picture of the actual profile page, so the bio, the grid and the follower ratio are what get judged. Being real and active is 30 percent of the grade on its own: a bought audience or an abandoned account scores badly whatever the headline number says.
Before commenting, the agent looks at the post image, describes what is in it, and writes about something in that description. If it cannot see the post, no comment is left, and there is no falling back to a generic compliment. Your pitch never reaches the comment writer, and a draft that sells or repeats itself is stopped before it goes out.
Outreach runs from your own connected Instagram or LinkedIn account, under an hourly and a daily ceiling for each kind of action, with randomised gaps, quiet hours, and a brake that pauses everything if activity spikes.
Research and outreach run on Stockholm-hosted, EU-resident infrastructure, consistent with GDPR.
Influencer work doesn't stand alone. The same agents run your social media, handle the outreach to brands and partners, and plan the wider marketing, all in your voice and on one platform.
Two ways, both yours to steer: the words and hashtags of your niche, and the followers of competitor or aspirational accounts you name. Your own account is always excluded.
Against the fit description you wrote. A picture of each profile is graded on fit, on whether the account is real and active, and on how likely they are to engage, and comes back as a score out of ten with a letter grade and one or two sentences saying why. Only creators clearing the score you set are contacted, seven out of ten by default.
No. Each message is written from that person's own profile and sent in their language, kept to a couple of sentences, with explicit rules against flattery openers, hype and more than one emoji. Your sign-off is used only where it fits.
Instagram and LinkedIn for outreach, from your own connected accounts. Facebook is supported for posting, not for outreach.
An hourly and a daily cap on each kind of action, around 36 hours between two actions on the same person, randomised gaps, quiet hours, and a brake on bursts. And a comment it cannot ground in the actual post is skipped rather than filled with something generic.
Invite-only while we onboard a small number of businesses. Request an invite and we'll be in touch.