Skip to main content

Upfluence MCP: Use cases and example prompts

The Upfluence MCP connects Claude, ChatGPT and other AI assistants to your Upfluence workspace, so you can run campaigns from a chat instead of clicking through the apps. Below are use cases our team and clients run today, each with a prompt you can copy and adapt.

Before you start

Connect the Upfluence MCP to your AI assistant first. Set up the Upfluence MCP

Four things that make every prompt below work better:

  • Name the campaign exactly as it appears in Upfluence, or paste its link. The assistant matches on the name and will stop if it finds none or several.

  • Say what the assistant may and may not do. The MCP can both read your data and act on it: send emails, change creator statuses. When you only want information back, write it in: only look, never send emails, reply to anyone or change a creator's status.

  • Two kinds of email, two behaviours. A custom reply or a first-contact email is written by the assistant and shown to you in full before anything leaves: you reply send to approve it. A campaign template (invitation, shipping, drafting, publishing, payment) can't be previewed, so the assistant tells you which template is going to whom and waits for your go-ahead instead. Either way the email goes out from your own address through Upfluence, not from your mailbox.

  • Drop the @ from handles. Creator lookup fails on @username and succeeds on username. Applicant lists usually arrive with the @ attached, so tell the assistant to strip it.

Scheduled tasks run in your assistant, not in Upfluence. They are available on Claude paid plans.

Tips: Set them to run in the cloud rather than on your own computer: a task tied to your machine won't fire once the laptop is closed, which is exactly when an overnight or early-morning run is meant to happen.

Your choice of model matters for the longer prompts below. Anything that works through a whole campaign takes many steps in a row, so switch to the most powerful model your plan includes before you run it. Shorter questions are fine on any of them.

Your daily workflow

Get a daily briefing of creator replies and applications

Use case. Your campaigns already track every reply, application and counter-offer. This brings that picture to you first thing, across all of them at once: who replied, who applied, who countered and at what price, in one prioritized list. Set it up once as a scheduled task and it arrives every weekday before you start.

Example prompt

Using Upfluence, set up a scheduled task that runs every weekday at 8:50 AM Eastern.

Only look, never send emails, reply to anyone or change a creator's status.

Check all my active campaigns (ignore archived ones) for anything new since my last working day ended at 6 PM Eastern (on Mondays, since Friday 6 PM), up until the moment it runs:

1. Creators who replied to my emails and I haven't read yet.

2. Creators who applied or sent a counter-offer (the "Influencer responded" and "Counter-offer received" stages).

Give me a short list for each, grouped by campaign, one line per creator with their name, what happened (or the time of their reply) and, for counter-offers, the price they're asking. Each creator links to their campaign.

Replies that don't belong to any campaign go in a separate "Direct conversations" group with the creator's name and email, no link.

Start with one line summing it up, like "Since Tue 6 PM -> Wed 8:50 AM ET: 3 unread replies, 2 applications". If a section has nothing, write "Nothing new."

If something couldn't be checked, tell me instead of skipping it. Don't guess anything.

If you find anything, send me a notification.

Reply, advance and reject from the chat

Use case. Once the briefing tells you what needs attention, handle it in the same place. Replies come back as a draft first, so nothing leaves your account unreviewed. Status changes can be done one creator at a time or in bulk across several campaigns.

Example prompts

Using Upfluence, for campaign [name or link], reply to creator [name] saying that we'll get back to him/her in a bit.

The assistant drafts the email and shows it to you. Reply send to send it.

Using Upfluence, for campaign [name or link], advance creator [name] to the next stage.

Or

Using Upfluence, for campaign [name or link], reject creator [name].

Three things to know before you work this way:

  • Creators move one stage at a time. Campaigns run stage by stage, in Upfluence as much as in the chat, so the AI assistant moves a creator to the next stage and no further. Nobody jumps straight from applying to shipping either way.

  • Rejecting only works before a creator is approved. Up to that point you can reject from the chat. Once they're approved and into the collaboration, reject them in the campaign instead.

  • Moving a creator to the next step doesn't send that step's email on its own. The email only goes out if you ask for it: "Using Upfluence, for campaign [name or link], send the shipping email to [creator names]". Same for the drafting, publishing and payment emails.

Creator sourcing and outreach

Find creators and invite them in one go

Use case. Searching, shortlisting and first contact are three steps you'd normally take one at a time. When you already know exactly who you're after, you can describe the whole sequence in one go: the assistant searches the same creator database, adds the matches to your campaign, and sends your invitation template. You still review the result in the campaign afterwards.

Example prompts

Find me 20 creators [criteria] and add them directly to my [campaign name] campaign. Once in the campaign, send the invitation email to all creators.

Or

Using Upfluence, set up a scheduled task for 9:00 AM EST today: send the invitation email to every creator at "Ready to invite" in my [campaign name] campaign. List who it will go to before it runs, and skip anyone who has already been invited or already has an email thread.

Tip. Ask it to leave out anyone you already have: exclude creators already in my [list or campaign name]. A whole list or campaign can be ruled out in one go, which stops a second round of sourcing handing you the same people back.

When your creators are in another time zone

Use case. Teams recruiting from Australia, Japan or Korea into the US run into the same thing: the invitations land in the middle of the recipient's night. Two details make a scheduled invite arrive where you want it.

  • Upfluence spaces campaign emails out. There are about 3 minutes 45 seconds between each one, so a batch of a hundred takes most of a working day to finish sending. Tell the assistant about the gap and the window you want to send in, and it works out how many creators it can invite per run.

  • Make it a project rather than a one-off prompt if you're recruiting continuously. Put the campaign, your creator criteria, your sending window and your daily cap into the project instructions once. You keep adding creators through the week, and the scheduled task invites the right number of them at the right hour each day.

Example project instructions

Here is how I work on the [campaign name] campaign.

- My goal is to recruit [100] new creators a month.

- Invitations may only go out between [9 AM and 5 PM Eastern], Monday to Friday.

- Upfluence sends campaign emails about 3 minutes 45 seconds apart. Work out how many invitations fit in that window and never queue more than that in one day.

- The creators I'm looking for: [criteria].

This is ongoing, so I'll keep adding creators to the campaign. Each day, check who is waiting at "Ready to invite" and send the invitation email to that day's quota, oldest first. Tell me afterwards who it went to. Skip anyone already invited or already in an email thread.

When you've finished sourcing, ask me whether I want the list exported or the creators added straight to the campaign.

Save that as your project instructions, then ask for the daily run as a scheduled task in the cloud.

Check a creator before you reach out

Use case. The audience breakdown and brand history sit on every creator's profile. Pulling them into the chat is useful when you're weighing several creators against each other, or when you want the verdict in the same place you're about to draft the outreach.

Example prompt

Using Upfluence, look up [handle, with no @ in front]. Show me the audience age, gender and country breakdown, their engagement rate, and the brands they've worked with. Finish with a one-line verdict on whether they fit a [niche] campaign targeting [country].

If any of this data isn't available, say so instead of estimating it.

Write a personalized line for every creator in a bulk mailing

Use case. Bulk mailings convert better when each one opens with something real about that creator. The assistant pulls the campaign roster from Upfluence, researches each creator, writes one personal sentence each, and hands you a CSV you import back into Upfluence as a custom merge field. You then drop {{personal_line}} and/or {{personal_subject_line}} into your email template.

The MCP returns profiles, bios, locations, platform stats and brand mentions, but not the content of recent posts. To reference a specific recent post, install the Claude in Chrome extension and keep it active: the assistant opens each creator's public profile and reads the captions itself. On YouTube it can read the video transcripts too, which is where the most specific lines tend to come from.

Example prompt

I want to create personalized merge fields for a bulk mailing in Upfluence. Please:

- Use the Upfluence MCP to find my campaign [campaign name] and pull all creators with the status "Ready to invite". Then, get their profiles (bio, location, platforms, email) and their brand mentions.

- Use Claude in Chrome to open each creator's largest social profile ([Instagram / TikTok / YouTube]) and look at their 3–5 most recent posts and their bio. Find one specific, recent, positive detail about each creator, like a trip, a project, a series, a milestone or a standout post. Read all captions on these posts thoroughly. If the largest platform is YouTube, read the video transcripts as well.

- Write one personalized sentence per creator that I can drop into this email: "Hi {{first_name}}, I hope you're doing well. {{personal_line}}. We think you'd be an amazing fit for our campaign..."

- Make it warm, grammatically correct, casual, and specific with no emojis.

Only use details you actually saw. Don't guess or invent anything.

- Avoid sensitive or personal topics (health, relationships, politics, family).

- If you can't find a solid detail, write a generic fallback line and flag it.

Write one personalized email subject line per creator that I can use as {{subject_line}}.

- Base it on the same verified detail as their personal_line, but word it differently so the subject and the first sentence don't repeat each other.

- Keep it short: under 50 characters (about 4 to 8 words), in sentence case.

Make it plain, specific and honest about what the email is. It should still make sense after they open it.

- It should read like a note from a real person, not a newsletter or an ad.

To avoid spam filters: no ALL CAPS words, no emojis, no exclamation points, no "$" or "%", and no spam-trigger words like "free," "exclusive," "opportunity," "selected," "guaranteed," "act now," "limited time," "urgent," "win," "deal," "collab alert".

- No fake reply or forward prefixes (Re:, Fwd:) and no false familiarity ("Following up," "As promised," "Our chat").

- No vague curiosity bait ("Quick question," "You'll want to see this," "Can I ask you something?," "Big news") and don't use only the creator's first name as the subject.

- The same rules apply as the personal line: only use details you actually saw, nothing sensitive or personal.

- If there's no solid detail, use a simple fallback and flag it.

Good examples (specific, plain, not click-bait):

  • Your Lisbon pastry ranking

  • Loved the drugstore dupes series

  • An idea for your GRWM Fridays

  • Saw your Seoul skincare haul

  • Your 30-day curly hair diary

  • Congrats on hitting 100 videos

  • Your salon vs. at-home color test

Good fallback examples (when there's no solid detail):

  • Campaign invite

  • A campaign we think you'll like

Bad examples (don't write anything like these):

  • "Exclusive opportunity just for you!!" (spam words, punctuation stacking)

  • "You've been selected" (reads like a scam)

  • "FREE L'Oréal products inside" (caps, spam trigger)

  • "Quick question" (vague curiosity bait)

  • "Re: our collab" (fake reply)

  • "Your content is amazing" (generic flattery that could apply to anyone)

- Sound like a real person writing a quick note, not a marketer or an AI.

- Don't use phrases like: "I came across your profile," "your content really resonates," "I love your vibe," "elevate," "authentic," "in today's world," "game-changer," "obsessed". Do not only go based off of YouTube video titles if YouTube is the main platform. Actually read the captions and transcripts.

- No em dashes, no "AI speak", no exclamation-point stacking, no flattery that could apply to anyone.

- For the personal line, reference the specific thing plainly, e.g. "Your Lisbon pastry ranking had me taking notes" not "Your amazing travel content truly inspires!"

- Make a CSV with these columns: name, media_url, email, personal_line, personal_subject_line.

- Separately, show me in chat the source (the link to the post or profile you used), and confidence (high/low).

- Give me the preview table and the CSV at the same time.

- Do not send any emails or change any creator statuses.

The source and confidence columns matter: they let you spot-check where each line came from before the mailing goes out.

Tip.

  • Reading profiles in Chrome is slow going, since every creator means opening a page and reading it, so run this on a shortlist rather than a 200-creator campaign, or split it across a few sessions.

  • And give the merge field a name you can version, like personal_line_01: a line written today won't suit the same creator in a different campaign three months from now, and a fresh import creates a new field rather than updating the old one.

Find creators whose engaged audience is in a specific city

Use case. Every Instagram profile in Upfluence carries an engaged-audience breakdown right down to city level. Audience rules filter on country, so when you're planning an event or tracking store visits, the assistant is a fast way to put the city question to a whole shortlist at once and get them ranked.

City-level audience data is available for Instagram only.

Build your list first on who the creator is: niche, platform, and the wider region they live in. Then let the assistant do the audience pass. The location filter in Search tells you where the creator is, not where their followers are, and the two often diverge: a Miami creator can have most of their audience in Paris.

How to run it

Step 1: Export your creator list or campaign as a CSV, upload it to the chat, then:

Use the Upfluence connector to check which creators in this exported list have an engaged audience on Instagram from [city]. Pull each creator's Instagram audience city breakdown and rank them by estimated engaged followers in that city.

Flag any creators with no audience in that city, and note anything unreliable.

The list comes back ordered by engaged followers in that city rather than by follower count, so genuinely local creators rise to the top.

Step 2: Refine it, then act on it.

The first ranking is rarely the one you act on. Stay in the same chat and send whichever of these you need.

Start by widening the city, which matters more than it sounds. Suburbs are listed separately in the data. Ex. Santa Monica and Malibu don't count towards Los Angeles unless you say they should:

Include the surrounding metro area for [city]. For Los Angeles, count Santa Monica, Pasadena, Malibu and West Hollywood as well, and treat "SF" as San Francisco.

If reach in the city matters more to you than engagement, change what the ranking rewards:

Re-rank by city share only, ignoring engagement.

Tip. Read the reliability notes that come with the ranking. They tell you how large a sample each percentage rests on, and they flag accounts whose numbers don't add up, such as a big following in the city with almost no likes.

Before you reach out, make sure these creators are new to you:

Check whether we've already emailed any of these creators.

Then move the ones you want into a list or a campaign, and carry on from there as usual:

Add the top 3 creators to [list or campaign name].

Once a few creators have worked out, keeps the search running from the ones you've already approved:

Find me more creators like these.

Campaign operations

Filter ambassador and affiliate applicants

Use case. When your ambassador or affiliate programme is open to applications, you end up checking the same handful of things on every single applicant: follower count, engagement rate, where their audience is based, what they post about. Tell the assistant those requirements once and it checks every applicant against them, then hands you a shortlist with the reason behind each decision. Nobody is added to a list until you confirm, and your original applicant list stays untouched.

Applicants reach you one of two ways, and both work:

  • They applied through an Upfluence landing page, or came in through a Live capture matching and are already saved to a list. Export that list as a CSV and attach it to the chat, or paste the handles.

  • They applied through your own site or an affiliate app (GoAff Pro, Refersion, Okendo and similar). Connect your inbox and the assistant reads the handles out of the notification emails.

Example prompt

I'm filtering ambassador applicants for [brand].

Criteria, all mandatory:

- At least [10,000] followers on [Instagram / TikTok]

- Engagement rate above [2%]

- At least [60%] of their audience in [country]

- Content in the [niche] space

Here are the applicants: [paste the handles, or attach the CSV export of the "[source list name]" list].

For each one, pull their followers, engagement rate, audience location breakdown and content categories from Upfluence, then tell me whether they pass. If a metric isn't available (private account, no data), flag them for manual review rather than rejecting them. If a handle isn't found in Upfluence at all, say so and skip it.

Show me a table with handle, platform, followers, engagement rate, audience location, niche, pass or fail, and the reason for any failure. Don't add anyone to a list until I've confirmed the batch.

Once I confirm, add everyone who passed to my "[destination list name]" list.

If you want a copy of the shortlist outside Upfluence as well, ask for a ready-to-import CSV with profile_name, media_url and email. That format imports straight back into a list.

Tip. The assistant looks each applicant up individually, so a few dozen come back quickly while several hundred make for a long run. Split a large export into batches: each one finishes sooner and the table is easier to check over.

Optional: email the applicants who qualified. They're waiting to hear back, so you can confirm it in the same chat right after the shortlist is added:

Send everyone I just added to "[destination list name]" a short email letting them know their application was successful and they're now part of the [brand] programme. Tell them what happens next: [what happens next]. Show me the draft before you send anything.

The assistant writes it once, you approve it, and it goes to the whole batch from your own address. Ask for a separate note to the ones who didn't qualify if you want to close the loop with them too.

Make sure no creator is left waiting for a reply

Use case. Replying quickly keeps creators engaged, but on a busy campaign some replies slip. You can let your AI assistant check the campaign on a schedule, find anyone still owed a reply, and draft it for you to approve.

The assistant works to a schedule you set rather than reacting the second a status changes, so pick a cadence that suits the campaign: hourly through a launch week, daily once it's running steadily. Each run it checks the campaign, finds everyone in each situation and drafts the replies for you to approve.

Example prompt

Using Upfluence, set up a scheduled task that runs every weekday at 9:00 AM [timezone] for campaign [campaign name].

On each run, check for three situations and handle them in this order:

1. Creators who reached "Influencer responded" or "Counter-offer received" and haven't been acknowledged yet. Reply in their existing thread thanking them for taking the time, and let them know we'll be in touch within 3 working days.

2. Creators at "Influencer rejected" who haven't had a closing note. Reply thanking them, saying we had more applications than expected this time and we'll keep them in mind for the next campaign.

3. Creators whose shipping email went out 5 or more days ago with no reply since. Reply asking whether the product arrived and saying we're looking forward to seeing their content.

Before sending anything, show me every email you plan to send, grouped by situation, with the creator's name and the date of the last email in their thread. I'll reply "send" to approve.

Never contact the same creator twice for the same reason: check the thread history first. If a creator has no email thread yet, list them for me instead of starting one.

Three things worth knowing about this one:

  • These are ordinary replies, not campaign templates, so you can word them however you like. The five templates stay for the campaign steps they belong to.

  • You don't have to keep track of shipping dates yourself. Each creator's email thread records when every email was sent and whether they've answered since, so the assistant can work out who passed the five-day mark without replying.

  • Each situation is a separate instruction. Drop the ones you don't want rather than rewriting the whole prompt.

Track drafts waiting on your review

Use case. Your campaign already knows which drafts are waiting on you. This carries that into the task manager your team works from, so a pending draft sits alongside everything else you owe someone and anything waiting too long gets flagged. It reads Upfluence and writes nothing back: no approvals, no emails, just the tasks and a digest.

Tip. Name the campaigns you review rather than asking for all of them. Your campaign list comes back fifty at a time and most workspaces hold far more archived campaigns than live ones, so "all my active campaigns" spends several rounds finding them before it reads a single draft. Two or three names is quicker and gets you the same list.

Before you start. This one reaches beyond Upfluence, so connect two more things to your assistant first: the task manager your team uses, and the channel you want the digest in. It can only create tasks and send messages in tools it's connected to.

Example prompt

Sync Upfluence drafts awaiting review into [task manager], then send me a digest on [Slack / email].

CAMPAIGN NAME: [campaign names, or "all my active campaigns"]

TIMEZONE: [your timezone]

FLAG AS PRIORITY AFTER: [X] days

CHASE AFTER: [X] days

SCOPE: read only on Upfluence. Never approve, reject, shortlist a creator or send any email. If the data suggests one of those is warranted, mention it in the digest and stop there.

1. Find the non-archived campaigns named above, or all of them if I said all. If a name I gave matches none or several campaigns, tell me the candidates and stop.

2. Pull every contribution in those campaigns with the status "drafts review", paging through if there are more than 100. If there are none at all, tell me "no drafts pending review" and stop without creating or touching a single task.

3. Pass every creator ID from those contributions in one lookup, up to 100 at a time, and keep each creator's name, Instagram or TikTok username, follower count and location. Note which ones have had a draft rejected before: those are resubmissions waiting on a second look, not first reviews.

4. In [task manager], keep all these tasks together in one place called "Creator draft reviews" — create it if it isn't there. Check it first for an open task containing the contribution ID and skip any draft already tracked, so nothing is duplicated. For the rest, create one task per draft: title "Review draft - @username [contribution_id]", or "Re-review draft - @username [contribution_id]" for a resubmission. If a creator has no Instagram or TikTok handle, use their name in place of the username. Put the creator's details and the campaign name in the description, due today, and mark anything last updated longer ago than the priority threshold above as high priority.

5. Every task and every digest line links to the creator's panel in the campaign:

Build it from the IDs you already have. Never use the links that come back on the contribution itself — those are the creator's own upload and brief pages, not mine.

6. Send me a digest grouped by campaign: one line per new task with resubmissions marked, a count of the ones already tracked, and a separate "Waiting longest" list of anything last updated longer ago than the chase threshold.

If any step errors, tell me, make no further writes, and stop.

Paste it into a chat to run it once. To have it keep watch on its own, ask for it as a scheduled task "run this every weekday at 9am" and set the task to run in the cloud, or it won't fire on a morning your laptop stays shut.

Tip. Do a dry run before you schedule it. Run the prompt by hand once, with step 4 (the step that creates the tasks) changed to "list the tasks you would create, but don’t create them". You get to read the titles and descriptions before a few hundred real tasks land in a shared project.

Monitor gifting spend on a campaign

Use case. Gifting orders and their values live on the creator who received them. This gathers them into one spreadsheet: what you have spent across the campaign, what each creator was sent, and which items creators pick most. It is also the list you work from if you fulfil gifting by hand, because it carries every recipient and address.

What comes back. Each creator's gifting arrives with the order number, the date it was placed, whether it is an initial order or an extra, its status, every product with quantity and unit price, the recipient's name and shipping address, and a link to the order in your connected shop.

Tip. Ask for the split between initial orders and extras. The initial order is what you meant to send and an extra is a top-up ordered later, so the two side by side show you how much of the spend went beyond the plan.

Example prompt

  • The quick version

Using Upfluence, what did I spend on gifting for campaign [campaign link]? Add up the product value of every order that isn't cancelled, and tell me how many orders, how many items and how many creators that covers.

  • The full report. When the figure is going to a client or into a monthly review, ask for the file.

Build me a gifting spend report for campaign [campaign link], covering orders placed in [date range]. State the date range at the top.

Go through the whole campaign and keep only the creators who have gifting orders. For each of those, take every order: order number, date placed, initial order or extra, status, and each product with its quantity and unit price, plus the recipient's name and shipping address. Fetch the creators' names and emails in one lookup. Give me the amounts in [your currency]. Where a field is empty, write "not shown" and never fill one in yourself.

Give me one .xlsx with two tabs, using live formulas for the totals so they stay right if I edit the file.

Tab 1, "Gifting spend": the total excluding cancelled orders at the top, with the order, item and creator counts beside it. Then that total split by status, and again by initial versus extra. Then one row per product per order: order number, date, creator, email, recipient, product, quantity, unit price, line total, initial or extra, status, and "counts toward spend?" set to no for anything cancelled.

Tab 2, "Item breakdown": one row per product with items sent, how many creators received it and its value excluding cancelled orders, sorted by items sent, with a total row. Then the most-sent item, the average items per order and the largest single order.

Two things to raise in the chat rather than bury in the file: any order that looks like a test, meaning a placeholder name or address, and any creator sent the same product twice within a few days, which is usually a duplicate rather than a second gift.

Before you send it, check that the status split and the product table each add up to the headline total. Reply with the file, the headline total, the result of that check, and anything you had to decide for yourself.

The totals in the file are formulas rather than typed numbers, so if you strike a row or correct a quantity the headline follows. That matters when you have to adjust something before sending it on.

Tip. The assistant goes through the campaign and keeps the creators who have gifting orders, so on a big campaign give it a moment before the first number appears. Narrowing it by date range is quicker when all you want is the current month.

Keep it current. Orders move from pending to fulfilled or cancelled after the report is built, so rather than making a new file each time, send this into the same chat whenever you want fresh numbers:

Update my gifting spend report for campaign [campaign link]. Read the gifting orders again the same way. Keep one row per product per order, add a row for every new order, and update any row whose status, quantity or price has changed. Never delete a row: cancelled orders stay, marked no under counts toward spend. Keep the totals as live formulas. Reply with the headline total first, then the orders you added, the orders you updated, and anything unclear.

Point the same prompt at a Google Sheet instead of a file if you want one living document your team can open, and the sheet then keeps a history you can follow an order through. You can also ask for the refresh as a weekly scheduled task, and just read the summary when it arrives.

Sharing it. Upload the file to Google Drive, open it in Sheets and share from there. If you would rather send a page than a file, ask for the report as a shareable page in the same chat. It is a snapshot, so ask again when the numbers move, and who can open it depends on your plan and sharing settings.

Tip. Read the list before you send it to a client. Test orders with placeholder names and addresses sit in the same gifting history as the real ones and get counted, and so do duplicates: the same creator receiving the same item twice within a few days is usually an ordering mistake rather than a second gift. The prompt flags both, but look at them yourself before the number goes anywhere.

Track outreach results across your campaigns

Use case. You want a cross-campaign overview of outreach performance: how many invitations went out, how many creators applied, made a counter-offer, got rejected.

Example prompt

Using the Upfluence MCP, I want to track email response rates across my campaigns [campaign name]: how many emails were sent and how many creators responded, rejected, and so on.

Claude will pull the data across your campaigns and calculate response rates for you. It will also work out on its own how to define a responded creator, meaning anyone who moved past Awaiting application. If you're using a different AI model and it needs guidance, add this line to your prompt:

Count a creator as responded once they moved past Awaiting application, whether they applied, replied, countered, or progressed further.

Tip. Make sure your prompt mentions the Upfluence MCP, or send it in a conversation where that context is already set. Without it, the AI may try to answer from general knowledge instead of pulling your live campaign data.

Write your own prompts

Your campaigns won't look exactly like the ones above, so the real value is in writing prompts for your own routine. The prompts here are long on purpose, and the length is what makes them reliable. These are the patterns to carry into your own:

  • Set the scope before the task. A line like only look, never send emails or change a creator's status at the top of the prompt keeps a research run from turning into an action run.

  • Say what a number includes. "Total gifting spend" is ambiguous; "total price of all created gifting orders" is not.

  • Ask for a preview before a file. A table in the chat lets you catch a bad line before you import 200 of them.

  • Tell it what to do when it is unsure. "If you can't find a solid detail, write a generic fallback and flag it and tell me instead of skipping it" are what stop an assistant from quietly inventing something.

  • Ask where each answer came from. A source column or a link per creator makes the output checkable.

  • Spell out the failure case. "If none or several campaigns match, tell me the candidates and stop" saves you from a run against the wrong campaign.

  • Reuse your prompts. Keep the ones that work in a project or a note, with the campaign name as the only thing you swap.

What to watch out for

What you'll run into

What to do

Creator lookup fails when the handle starts with @

Pass the handle without the @

A campaign's creators can be pulled in full, a page at a time; the creators sitting in a list can't be fetched the same way yet

For a list, export it as a CSV from Upfluence and attach it to the chat, or paste the handles

Creators move one step at a time, and rejecting only works before a creator is approved

Move them one step at a time, the same way the campaign screen works. If an action isn't available, ask which step the creator needs to be at for it

Asking to move a creator to the next step doesn't send that step's email

Ask for it separately: Using Upfluence, for campaign [name or link], send the shipping email to [creator names].

Recent post content isn't returned, only profiles, stats and brand mentions

Use the Claude in Chrome extension to read captions on the creator's public profile

Nothing notifies the assistant when a creator's status changes

Use a scheduled task that checks on a cadence you choose

A scheduled task tied to your own computer won't run when the laptop is closed

Set scheduled tasks to run in the cloud

Audience data by city covers Instagram only

Use country-level audience rules for other platforms

Re-uploading a CSV creates new merge fields rather than updating existing ones

Name your fields clearly on first import

If there's something you'd like the assistant to reach, send us a feature request. We're actively adding to what the assistant can reach.

Did this answer your question?