Give the agent the operational context
State the product, program, target creator, platform, region, evidence requirements, and exclusions.
Premium MCP access
Run UGC creator sourcing from the LLM workspace your team already uses.
Premium users receive a private Streamable HTTP MCP URL. Connect it to compatible clients and call the same server-side creator-sourcing tools used by the web application, with plan enforcement and saved user preferences.
Private per-user MCP URLCreator lists open in the web appSame server-side sourcing logic
Workflow
From request to usable output.
Inspect the structured output
Review sources, score reasons, classifications, risk flags, and any missing evidence before acting.
Keep the human decision
Use automation for the first pass while your team controls outreach, approvals, commercial terms, and publishing.
01
What the MCP exposes
The MCP workflow can research companies, find proof-backed brand collaborators, analyze seed creators, discover profiles, enrich and score them, list or remove leads, export CSV files, analyze creator content, research trends, ideate content, draft scripts, and review UGC applications.
02
Where it works
The product includes setup guidance for Codex, Claude Code, ChatGPT workspaces that permit custom connectors, Claude, Cursor, Windsurf, Cline, Roo Code, Continue, Hermes, OpenClaw, and other compatible Streamable HTTP clients. Client capabilities and setup requirements can change, so use the linked official setup documentation.
03
Security and entitlement
The URL is tied to the signed-in user and is shown only to eligible Premium accounts. API keys for Instagram, TikTok, Gemini, Firebase Admin, and Stripe stay on the server and are not delivered to the LLM client.
04
Use the web app for review
MCP-generated creator lists can be opened in UGC Agent for a readable table, filtering, saved-list organization, and gated export. This gives operators a review surface instead of leaving creator data buried in a chat transcript.
05
A complete MCP sourcing turn
A complete request can research the product or company context, translate the goal into search angles, discover Instagram and TikTok candidates, enrich and score the public profiles, retain qualified leads, and return an app URL for table review and CSV export.
For manual control, compatible clients can call discovery and enrichment tools separately, refine the active list, remove leads, or export after the requested quality and count are reached. Tool availability remains subject to the active plan and server configuration.
06
Persistent lists and preferences
Signed-in MCP runs can reuse saved creator preferences, user exclusions, deduplication history, and durable creator-list identifiers. Reusing the returned list ID keeps follow-up requests on the same list instead of silently starting an unrelated result set.
A one-run preference override does not change the saved global preference. When requirements conflict, the operator should decide whether to use stored constraints, a temporary override, or no saved preference for that run.
07
Review generated campaign scaffolding
MCP tools can also research current content patterns, analyze retained creators' public posts, ideate formats, and draft script scaffolds. Those outputs need product-specific rewriting and claim review before they become creator direction.
Do not present a generated trend, hook, script, or calculator result as a customer outcome. Keep sources with research outputs and open the web application when a table, saved list, or export is easier to inspect than a long chat response.
Treat the private MCP URL like an account credential. Do not publish it in documentation, screenshots, public repositories, or shared prompts. If access changes, use the product's current setup and entitlement flow rather than copying a connection from another user or workspace.
Confirm the client supports the required Streamable HTTP connection before treating a setup example as current compatibility evidence.
Questions
What teams usually need to know.
Is this feature available in the web app?
Yes. Availability and usage limits depend on the active UGC Agent plan.
Does UGC Agent make the final creator decision?
No. It provides structured evidence and a first-pass recommendation for human review.
Does the feature use private social data?
The sourcing workflow uses public creator profile and content signals supplied by configured data providers.
Can I export the result?
Creator leads and unlocked analysis fields can be exported where the workflow supports CSV export.
Can the same workflow run through MCP?
Premium MCP access exposes the supported server-side tools to compatible LLM clients.
Turn the brief into a creator list.
Try a restricted sourcing run free, then upgrade only when you need more visible or monthly creator leads.