PurrPlan

AIエージェントが操作できるソーシャルメディアスケジューラー：リモートMCP経由で12以上のネットワークにわたり、計画、公開、受信トレイ、分析を実行。

GitHub
1
このMCPを試すスポンサー

ドキュメント

PurrPlan MCP Server

Official remote MCP server for PurrPlan — the social media scheduler your AI agent can drive. Plan, publish and analyze content across 12+ networks (LinkedIn, X, Instagram, TikTok, Facebook, YouTube, Threads, Bluesky, Pinterest, Telegram, Mastodon…) straight from Claude, Cursor, OpenClaw or your own scripts.

  • Endpoint: https://app.purrplan.ai/api/mcp (Streamable HTTP, MCP protocol 2025-06-18)
  • Health check (no auth): https://app.purrplan.ai/api/mcp/health
  • Auth: Bearer token (created in-app) or OAuth 2.0 with Dynamic Client Registration + PKCE
  • Docs: https://purrplan.ai/developpeurs (FR) · https://purrplan.ai/en/developpeurs (EN)
  • Partner API + webhooks: docs/partner-api.md — create a client, get a token, connect Facebook back to your tool, verify X-Signature. Worked example: examples/partner-flow.mjs.
  • Free library (this repo, MIT): a thin JSON-RPC client for that endpoint, plus the webhook signature check. Not a paid SDK. You supply the token.

Free library

# from a checkout of this repo
node --experimental-strip-types -e "
import { createPurrPlanClient } from './src/index.ts';
const client = createPurrPlanClient({ token: process.env.PURRPLAN_TOKEN });
console.log(await client.listWorkspaces());
"

createPurrPlanClient({ token, endpoint? }) posts tools/call to https://app.purrplan.ai/api/mcp only. It refuses a missing token, a reply without confirm: true, and a media URL that points at a private address. It does not call /mixpost/api. The 18 methods match the SaaS registry (src/tools.ts).

Quick start

Create a token in the workspace API & MCP page (https://app.purrplan.ai/{workspace_uuid}/api-mcp), pick your scopes, then:

Claude Code

claude mcp add --transport http purrplan https://app.purrplan.ai/api/mcp \
  --header "Authorization: Bearer YOUR_TOKEN"

Generic JSON (Cursor, Windsurf, OpenClaw, …)

{
  "mcpServers": {
    "purrplan": {
      "type": "http",
      "url": "https://app.purrplan.ai/api/mcp",
      "headers": { "Authorization": "Bearer YOUR_TOKEN" }
    }
  }
}

Clients without native HTTP support can bridge through mcp-remote:

{
  "mcpServers": {
    "purrplan": {
      "command": "npx",
      "args": ["mcp-remote", "https://app.purrplan.ai/api/mcp",
               "--header", "Authorization: Bearer YOUR_TOKEN"]
    }
  }
}

Tools (18)

ToolScopeWhat it does
list_workspacesreadWorkspaces you can act on
list_accountsreadConnected social accounts and their auth status
list_postsreadPosts with per-network versions, filterable by status
get_postreadFull detail of one post (versions, media, accounts)
generate_ai_textaiGenerate copy with the workspace's brand voice
create_draft_postwriteCreate a draft or scheduled post; array content = thread (X, Threads, Bluesky, Mastodon) or first comment (Facebook, Instagram)
create_storieswriteSchedule a cascade of stories (Instagram, Facebook), up to 30
update_draft_postwriteEdit content, accounts, media or schedule of a draft
delete_postwriteDelete drafts/scheduled posts (never published ones)
upload_media_from_urlmediaIngest an image/video from a URL (≤ 50 MB)
list_inboxinbox:readComments, DMs and mentions across networks
get_inbox_threadinbox:readFull thread of one conversation
manage_inbox_messagesinbox:readMark read/unread, archive, assign
refresh_inboxinbox:readForce an immediate inbox sync
reply_to_inbox_messageinbox:replySend a real reply as the connected account (confirm: true required, hourly cap)
get_analyticsanalytics:readFollowers, reach, impressions, engagement, clicks, per-network breakdown
get_top_postsanalytics:readBest posts + audience/engagement curves
get_post_statsanalytics:readStats of one post by uuid, per network (no top-50 limit)
plan_my_weekwrite+aiTurn a brief into N scheduled posts across the week

Scopes

read · write · ai · media · inbox:read · inbox:reply · analytics:read

Tokens carry only the scopes you grant. Inbox replying and analytics are deliberately separate scopes. Every tool call is audit-logged in-app.

Security

  • Untrusted content guard: inbox content is treated as data, never as instructions.
  • SSRF protection on all client-supplied URLs.
  • Rate limits: 100 req/min (API+MCP), 20/min (AI), 30/min (media).
  • Destructive/outbound actions (reply_to_inbox_message, scheduled sends) require explicit confirmation flags.

Pricing

API & MCP access is included in PurrPlan plans (from €9/month, MCP from the Pro plan) — not a paid add-on. 7-day free trial: https://app.purrplan.ai/app/register

Lifetime deal: pay €597 once, your agent posts forever — no subscription. Includes the Pro plan (API + MCP, AI credits and quotas). The only social scheduler with a lifetime option for agent-driven workflows.

Support