Mcp Creator Finder
Find influencers by keyword or niche across seven creator platforms, with follower counts.
- Transport
- Not stated
- Package
- —
- Registry id
- com.mambabuilt/mcp-creator-finder
No install snippet on purpose. A working MCP config is a command, its arguments and an environment block — the last two are where API keys live, so this catalogue never stores them and cannot publish them. Follow the link above for the authors' own instructions.
MCP server for the Mamba Labs Influencer Finder actor on Apify.
Keywords or a curated niche in, influencer handles out, one flat row per creator per platform.
Install
npx -y @mambalabsdev/mcp-creator-finder
Claude Desktop
{
"mcpServers": {
"mamba-creator-finder": {
"command": "npx",
"args": ["-y", "@mambalabsdev/mcp-creator-finder"],
"env": { "APIFY_TOKEN": "your-apify-token" }
}
}
}
Get an Apify token at console.apify.com/account/integrations.
Tool
find_influencers
Find influencers by keyword or niche across seven creator platforms, with follower counts.
| Input | Type | Required | Notes |
|---|---|---|---|
keywords |
array | no | Search phrases, one per line, for example "budget travel" or "meal prep coach". Each keyword is searched on each platform. Use niche instead to load a curated keyword set. |
niche |
string | no | Loads a curated set of search keywords for the niche (about 60 per niche, measured in the September 2026 pre-research). max_keywords_per_niche caps how many are used. Leave as custom to search only your keywords. Default "custom". |
max_keywords_per_niche |
integer | no | How many keywords from the niche set to search. 1 to 60. Default 10. |
platforms |
array | no | Which platforms to search, and which platforms a bare @handle is looked up on. A full profile URL carries its own platform and ignores this. Supported: TikTok, Instagram, YouTube, Pinterest, Twitch, Threads, and. |
max_creators_per_keyword |
integer | no | Cap per search. Search engines honor the site: filter for the first page or two only, so 20 to 30 per keyword with more keywords beats deep paging. Default 20. |
max_creators |
integer | no | Hard cap on rows returned, so a broad niche cannot run away. Default 200. |
follower_min |
integer | no | Drop creators whose follower count is known and below this. A creator whose count the search did not show is kept, so a later profile read can fill it. Default 5000. |
follower_max |
integer | no | Drop creators whose follower count is known and above this. Default 500000. |
us_only |
boolean | no | Launch scope is US creators. Discovery carries no country signal, so country_guess is null on every row of this actor and this field changes nothing here. It is accepted so the same input runs unchanged on the. Default true. |
batch_size |
integer | no | Rows fetched at once. Leave empty for the measured per platform default; the measurement is in the README. Higher is faster and, above the measured point, loses rows. |
contribute_to_shared_pool |
boolean | no | On by default. Contributes the public records this run finds to a shared creator and agency pool that all users of this actor read from. Only public data already in the returned rows is sent, and a contribution is not charged. Set false to read the pool and write nothing. Default true. |
Nothing is required. Influencer Finder answers a run with no usable input with a row carrying row_status and error_reason rather than failing, and the tool mirrors that.
Pricing
Influencer Finder is pay per event on Apify. Every price below is flat across the FREE, BRONZE, SILVER, and GOLD tiers.
| Event | Charged for | Price | Fires when |
|---|---|---|---|
actor-start |
Actor start | $0.001 | Once per run, on start. Covers the run overhead. |
creator-found |
Creator found | $0.007 | Once per creator row returned by keyword or niche discovery with a handle and a profile URL. A search that returns nothing charges nothing. |
Reading the output
Every row carries row_status and error_reason. A creator the actor could not read comes back as a row saying why, not as a gap in the list, so an absence is readable rather than inferred. Filter on row_status before loading a table.
From the project's README.