Find Influencers From ChatGPT, Claude or Gemini: Stellar's MCP Explained

Stellar now runs as an MCP server, which means the AI assistant your team already has open, whether that is ChatGPT, Claude or Gemini, can query it directly. MCP is the open standard for connecting AI tools to software, released by Anthropic and adopted since by OpenAI, Google, Microsoft and Amazon. On a normal Tuesday it looks like this: you ask for food creators with a German audience above 60%, and a vetted shortlist comes back with real audience data attached, inside the chat window you were already working in. It is included in every Stellar plan, with no add-on and no extra cost. The week of manual vetting that sits between a brief and a live campaign doesn’t get faster, it stops existing.

Two to Six Weeks, and Almost None of It Is the Campaign

Ask a marketing team how long it takes to get an influencer campaign live and the honest answer is somewhere between two and six weeks. Very little of that is creative work.

Here’s where the time actually goes:

• Hours scrolling Instagram and TikTok looking for the right profiles
• Vetting each one by hand, one browser tab at a time
• Shortlists that live in a spreadsheet and go stale within a fortnight
• Four tools that don’t talk to each other
• A result at the end that nobody can measure cleanly

None of these steps is difficult, which is precisely why they keep sliding down the priority list when a week gets busy. That’s how teams end up briefing creators they only half-know, from a shortlist somebody put together back in March.

The week you spend vetting is the week you don’t spend on the campaign, and that’s the trade an MCP connection takes off the table.

MCP Is Plumbing, and That’s Why It Matters

MCP stands for Model Context Protocol. Behind the acronym it’s a shared language that lets an AI assistant connect to a piece of software and actually do things inside it, rather than just describe what it would do if it could.

Before MCP existed, every AI tool needed its own hand-built bridge to every platform it wanted to reach, which is why so few of those bridges ever got built. Now there’s a single standard on both sides: software declares what it can do once, and any compatible assistant takes it from there.

The reason to pay attention is how quickly the industry lined up behind it. Anthropic released MCP as an open standard in November 2024, and four months later OpenAI added support across ChatGPT and its Agents SDK. Weeks after that, Google confirmed MCP support for Gemini, with Demis Hassabis describing it as the emerging standard for the agent era. By December 2025 the protocol had moved to a neutral foundation under the Linux Foundation, co-founded with OpenAI and Block, so no single vendor controls it. More than 10,000 public MCP servers are running today.

Standards wars usually drag on for a decade and end in an uneasy draw. This one took eighteen months and produced a clear winner.

Which is why this isn’t a chatbot bolted onto the side of Stellar. It’s Stellar itself, reachable from whichever assistant your company has already standardised on, with twelve years of creator data sitting behind it.

One Brief, One Chat, a Campaign Ready in 60 Seconds

The quickest way to understand it is to watch it. In our one-minute demo, the brief starts in Claude and everything it asks for happens live in Stellar.

Here’s what happens, step by step:

1. The brief. “We’re taking creators on a Paris trip. Find me German travel creators on TikTok.” The assistant asks who they should reach, and the answer is one sentence: mostly women, at least 70% of the audience in Germany, 100K to 250K followers, solid engagement.
2. The search. Stellar opens discovery on TikTok and sets every filter on its own: keywords, follower range, engagement rate, age, gender, country and language.
3. The vetting. It reads each match’s audience insights before answering, then comes back with the top five for the trip and the numbers behind each one.
4. The campaign. “Add them to a campaign and prep the shortlist for my client, with why each one fits Paris.” The campaign is created, five creators are shortlisted, and every profile carries a client-ready note on engagement, audience and fit.
5. The outreach. One last request, and a warm invitation to join the Paris trip is drafted and ready to send.

Nobody opened a filter panel, exported a spreadsheet or copied a number by hand. The video happens to use Claude, but the same conversation works in ChatGPT, Gemini or any other assistant that speaks MCP.

Ask It the Way You’d Ask a Colleague

Here’s what that looks like on a Tuesday morning.

Food creators with an audience at least 60% German. 50k to 300k followers. Engagement above 3%.

That single sentence is a complete instruction. Your assistant translates it into filters, Stellar runs the search across 440M+ profiles, and the shortlist comes back within seconds with the numbers already attached, so you never open the platform, never touch a filter panel and never export anything into a spreadsheet.

Then you keep going, because the first list is never the final one:

Drop anyone who posted for a competitor in the last six months.

Sort by engagement and give me the top eight with their audience split.

Each refinement costs you a sentence rather than another trip through the interface. Nothing about the underlying data has changed here, since Stellar still searches creators across Instagram, TikTok, YouTube and Twitch with 45+ filters exactly as it always has. What disappears is the distance between the question in your head and the answer on your screen.

The Answer That Matters Is Usually the Caveat

A shortlist is the easy part, because almost any tool can return names that technically match a brief. What separates a good shortlist from an expensive mistake is the second question, and that’s the one most teams stop asking somewhere around profile eleven.

Take a food creator with 790k followers on Instagram and another 600k on TikTok. She looks excellent on paper and would sail through a manual review without anyone raising a hand. Ask about her audience, though, and the picture changes quickly.

Her engagement rate runs between 0.5% and 0.8%, where a creator of that size in food would ideally sit closer to 2%. More importantly, only about a third of her Instagram following is actually in your target market, and on TikTok that drops to roughly a sixth.

She isn’t a bad creator by any measure. She’s a bad creator for a national campaign and a genuinely good one if what you’re buying is broad reach, and the gap between those two readings is worth several thousand euros to whoever signs the contract.

This is where talking to the platform earns its place. Asking “is her audience actually German?” costs four seconds and no context switch, so you ask it every time. Opening a new tab, finding the demographics view and reading through a breakdown costs three minutes, which is why by the eleventh profile you’ve quietly given up. Stellar shows audience demographics, fake-follower rates and past collaborations in either scenario. The only variable is whether anyone looks.

Good vetting was never really about access to data. It’s about how many questions you have the patience to ask.

Stop Opening Stellar and Just Ask It Instead

It’s 9:40 on a Monday and a client needs to know which creators in your portfolio are based in Paris, because a store opening got moved forward.

You have the answer already. Getting it out of the platform means a login, a portfolio view, a location filter, an export and a quick tidy-up before anyone else can read it, which comes to about eight minutes. That isn’t much, until you notice you actually did it at 11:00 rather than 9:41. Eight minutes of friction is enough to make almost anything wait. Ask instead, and the list arrives before you’ve finished reading the email.

The same logic applies to the jobs that never quite get scheduled. Forty profiles imported back in March, half missing a market tag, a dozen carrying follower counts from four months ago. Everyone knows the list needs cleaning and nobody has an afternoon spare, so it doesn’t happen. Tell Stellar to refresh the metrics, tag everything in the DACH region and flag any profile whose engagement has dropped, and it does the work and reports back on what it changed.

Once the platform answers rather than waits, a whole category of question becomes routine:

• Who in my portfolio is based in Paris and works in food?
• Which partnerships last quarter beat our engagement benchmark?
• Pull the full report for the spring launch: reach, impressions, EMV.
• Compare these two creators on audience overlap before I brief both.
• Refresh this list and tell me what changed.

That last group matters more than it first appears. Stellar already keeps every collaboration on record so you can go back to the creators who genuinely performed, and it already turns campaign activity into evidence in real time across reach, EMV and sales. None of that capability is new. What’s new is that the record answers when you speak to it.

A database you have to open is a reference. A database you can simply ask is closer to a colleague.

ChatGPT, Claude or Gemini: Bring the Assistant You Already Use

There is no Stellar assistant to learn. You connect the one your team already pays for, once, and keep working where you work.

• Any assistant that speaks MCP. ChatGPT, Claude, Gemini, Copilot and many more. If your company switches tomorrow, the connection follows.
• Included in every plan. No add-on, no extra cost, no separate contract.
• Your data stays yours. It is kept until you delete it, never shared with other clients and never used to train models.
• Your permissions, nothing more. An agent connected through MCP sees what your account sees and changes what your account can change. Read an instruction properly before you send it, exactly as you would before running a bulk edit by hand.

And if your team would rather pipe Stellar into its own dashboards than talk to it, the API is still there for that.

Conclusion

It’s tempting to file all of this under time saving, but that’s the smaller half of the story.

The hours you get back only matter because of where they end up going. Knowing which creator will say yes to a smaller fee because she actually likes the product. Noticing that someone’s audience has shifted, and having the conversation before the numbers force it. Sending the sample early, answering the DM, remembering what went wrong last time. That work is slow, human and largely unautomatable, and it’s the reason two brands with identical budgets end up with completely different results. Nobody ever won a partnership by filtering a spreadsheet faster.

Five things worth taking away:

1. MCP is an industry standard, not a Stellar feature. Anthropic built it, OpenAI, Google, Microsoft and Amazon adopted it, and it now sits under neutral governance at the Linux Foundation. Whether your company settles on ChatGPT, Claude or Gemini, this connection keeps working.
2. Discovery turns into a conversation. Food creators with a German audience above 60%, narrowed by follower band and engagement rate, becomes one sentence and two follow-ups rather than an afternoon of filtering.
3. Depth stops being expensive. When the second question costs four seconds instead of three minutes, people actually ask it, and the creator with a third of her audience in the wrong country gets caught before the contract rather than after the campaign.
4. Your portfolio starts answering back. The relationship history you spent years building is only worth something at the exact moment a decision is being made, and you can now reach it without leaving the email you’re halfway through writing.
5. The data stays reliable. Maintenance gets done when it costs a sentence instead of an afternoon, and current data is what makes every report downstream worth putting in front of a client.

None of this replaces judgement, taste, or the relationships you’ve built over years, and it isn’t meant to. What it clears away is the administrative layer sitting on top of them. If you want to see what your own week looks like without that layer, start with the platform and write down the first five questions you’d put to it. They’re probably the same five you’ve been putting off since March.

What would you ask Stellar if you never had to open it?

Connect ChatGPT, Claude or Gemini once, then find creators, interrogate their audiences, update your portfolio and pull your reporting entirely by conversation. Included in every plan. Thirty minutes is enough to see what changes.

Book a demo