Sign in

Agentswelcome

Creator Swarm was built agent-first. People use the app, agents use the same core through an API, and both work under the same rules.

  • REST API

    Every action in the app is an endpoint under /api/v1, described by an OpenAPI document.

  • MCP server

    Connect Claude or any MCP client over Streamable HTTP. Tools for reading, generating, publishing and admin.

  • TypeScript SDK

    A typed client generated from the OpenAPI document, with typed errors and paging helpers.

  • Webhooks

    Signed events when content is ready or fails, a post goes out or fails, or the spend limit pauses an item.

Connect from Claude Code (.mcp.json)
{
  "mcpServers": {
    "creator-swarm": {
      "type": "http",
      "url": "https://app.creatorswarm.ai/mcp",
      "headers": {
        "Authorization": "Bearer ${CREATOR_SWARM_API_KEY}"
      }
    }
  }
}
List Tier 1 reels with the SDK
import { createCreatorSwarmClient } from '@creator-swarm/sdk';

const cs = createCreatorSwarmClient({
  apiKey: process.env.CREATOR_SWARM_API_KEY!,
});

for await (const reel of cs.paginate('/reels', {
  params: { query: { tier: '1', limit: 100 } },
})) {
  console.log(reel.id, reel.score);
}

Built-in prompts

daily_run

Check spend, run discovery, pick the top 3 Tier 1 reels, create Format A content for each, and report back.

weekly_report

Summarise the last 7 days: posts, views, followers, cost per 1k views, best and worst post.

Keys with limits

read
See influencers, reels, content, posts, metrics and costs
generate
Run discovery, approve reels, create and regenerate content
publish
Approve, schedule and post
admin
Change settings, keys and webhooks

Paid tools stop at the daily spend limit. By default an agent can’t approve its own content for posting. Every call is written to the audit log with the key that made it.