Using pepe.business with AI agents via MCP
The Model Context Protocol (MCP) is the standard way for AI agents to interact with external services. pepe.business exposes a hosted MCP server with 15 tools.
What you can do
With the MCP server, your AI agent can:
It authenticates with the same pk_live_ API key the REST API uses — a key already in use works here too, and revoking it revokes the agent's access along with everything else that key could do.
Setup for Claude Desktop
Add this to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):
``json
{
"mcpServers": {
"pepe-business": {
"url": "https://pepe.business/api/mcp",
"transport": "http",
"auth": {
"type": "bearer",
"token": "pk_live_xxx"
}
}
}
}
`
Restart Claude. Now you can say:
> "Publish on Twitter and LinkedIn: 'We just shipped v2!' with the image https://example.com/v2.png"
Claude will call create_post automatically.
Setup for Cursor
- Auth: Bearer pk_live_xxx
Save and restart Setup for VS Code (Copilot)
Add to
settings.json:`json
{
"github.copilot.chat.mcp.servers": {
"pepe": {
"url": "https://pepe.business/api/mcp",
"headers": { "Authorization": "Bearer pk_live_xxx" }
}
}
}
`Available tools
The MCP server exposes 15 tools:
— the connected accounts a post can go to
get_account_status — plan, destinations used, current cost
create_post — publish now or schedule for later
list_posts / get_post — recent posts, or one in full with its analytics
retry_post — retry one that failed
generate_post_text / rewrite_text / translate_text — write or adapt copy
suggest_hashtags — with reach and competition estimates
analyze_sentiment — classify a message before replying to it
get_analytics — the last 30 days, totals and best posts
create_short_link / list_short_links — shorten and track a URL
run_autopilot` — read a website, generate the post, optionally schedule itSee the full tool list.
Try it now
That's it — your AI agent can run the board.