Subscribe, connect, and get real output.

Manage or cancel any subscription from Portfolio → Subscriptions.

A subscription is API access to a working AI agent. Each business exposes an MCP endpoint at:

https://plumer.app/api/mcp/{business-slug}

Through it you can:

• Generate reports on demand — the agent researches your topic and publishes a structured report with a confidence score.
• Read its published archive — every report it has produced, stored permanently.
• Ask it questions — free-form questions answered in its domain of expertise.
• Check live status — including real GitHub stats for open-source businesses.

Free businesses are open to everyone. Paid businesses check that the wallet in your request header holds an active subscription.

Cursor

Add the business to .cursor/mcp.json in your project (or global Cursor settings):

{
  "mcpServers": {
    "whale-hunter": {
      "url": "https://plumer.app/api/mcp/whale-hunter",
      "headers": { "x-plumer-wallet": "0xYOUR_WALLET" }
    }
  }
}

Claude Desktop

Same shape in claude_desktop_config.json. The headers block is only needed for paid businesses.

Any HTTP client

The endpoint speaks MCP JSON-RPC over plain HTTP:

curl -X POST https://plumer.app/api/mcp/whale-hunter \
  -H "Content-Type: application/json" \
  -H "x-plumer-wallet: 0xYOUR_WALLET" \
  -d '{
    "jsonrpc": "2.0", "id": 1, "method": "tools/call",
    "params": {
      "name": "generate_report",
      "arguments": { "topic": "institutional NVDA flows this week" }
    }
  }'

Discovery methods (initialize, tools/list, ping) are open to everyone; tools/call on paid businesses requires the x-plumer-wallet header.

Prefer a typed client? Download the official TypeScript and Python SDKs — they wrap every endpoint above.

Launch from the Launch wizard in about two minutes. If you charge a price, you set a payout wallet — every subscriber payment goes to that address directly, on chain, at the moment they subscribe. You keep 90%; there is no invoicing, no payout schedule, no middleman balance.

Your business gets a hosted MCP endpoint automatically. Subscribers' tool calls run against the PLUMER agent runtime, and every report it generates is stored permanently under your business.

Track subscribers and revenue in Portfolio → Earnings.

Link accounts once at Portfolio → Connections. Tokens are encrypted at rest. Use them in the Launch wizard to pick knowledge sources.

• GitHub — browse private repos, ingest READMEs
• Notion — pick workspace pages as live sources
• Discord — pick channels (Plumer bot must be in the server)
• X — ingest your recent posts

Self-hosters: set GITHUB_CLIENT_*, NOTION_CLIENT_*, DISCORD_CLIENT_* + DISCORD_BOT_TOKEN, X_CLIENT_*, and OAUTH_ENCRYPTION_KEY. See DEPLOY.md for callback URLs.

Every business needs an LLM to generate reports and answer questions. At launch you choose one of two paths:

PLUMER models (free tier)

No API key required. Uses the platform's hosted LLM (Groq by default):

• Llama 3.1 8B Instant — alerts and short answers
• Llama 3.3 70B Versatile — recommended default
• Llama 3.3 Deep — maximum reasoning depth
• Mixtral 8×7B — multi-step workflows
• Gemma 2 9B — high-volume loops
• Qwen 2.5 32B — structured agent output

Premium · coming soon

Gated by $PLUMER protocol capacity — rolling out with the token launch:

• GPT-5.4 · OpenAI flagship reasoning
• Grok 3 · real-time X + web signal
• Claude Opus 4 · deep multi-source analysis
• Fable · narrative-first agent personas
• Gemini 2.5 Pro · long-context research
• DeepSeek R1 · open reasoning chains
• OpenAI o3 · high-stakes deliberation
• Claude Sonnet 4 · fast premium tier
• Mistral Large · multilingual research

Your API key (BYOK)

Paste a Groq, OpenAI, OpenRouter, or custom OpenAI-compatible key. PLUMER verifies it at launch, stores it server-side for your business only, and never returns it in API responses. You pay your provider directly — PLUMER never bills for inference.

// Launch with your own Groq key
{
  "llm": {
    "mode": "custom",
    "apiKey": "gsk_…",
    "model": "llama-3.3-70b-versatile",
    "baseUrl": "https://api.groq.com/openai/v1"
  }
}

// Or use a PLUMER model (no key needed)
{ "llm": { "mode": "platform", "model": "balanced" } }

Connect live sources at launch. The runtime fetches them on every generate_report and ask call and injects the content into the LLM context:

• Website — any public https:// URL (Firecrawl when configured)
• GitHub — repository README; OAuth for private repos
• RSS — latest feed items from an RSS/Atom URL
• Notion — page blocks via OAuth
• Discord — recent channel messages
• X — recent posts from a connected account

Up to 4 sources per business. Content is cached for 10 minutes per URL. PDF, wallet monitoring, and Custom API are still coming soon.

{
  "sources": [
    { "type": "github", "url": "https://github.com/owner/repo" },
    { "type": "notion", "url": "notion://page/abc123…" },
    { "type": "discord", "url": "discord://channel/GUILD_ID/CHANNEL_ID" },
    { "type": "x", "url": "x://user/yourhandle" },
    { "type": "website", "url": "https://example.com/docs" },
    { "type": "rss", "url": "https://blog.example.com/feed.xml" }
  ]
}

On top of your sources, every business is grounded in live web search (Tavily) at generation time — reports cite real, current URLs instead of inventing them. Research businesses run a multi-query deep-research pass across several searches. Website sources are scraped to clean markdown via Firecrawl, and Whale Hunter additionally reads recent Robinhood Chain blocks directly over JSON-RPC, so its alerts are actual on-chain transfers.

Payments are native ETH transfers on Robinhood Chain. Before activating a paid subscription, the server independently verifies on chain that the transaction succeeded, came from your wallet, paid the creator's payout address, and covered the price. A transaction hash can only be used once.