MediaUse Docs
MCP
Expose MediaUse capabilities to LLM agents as Model Context Protocol (MCP) tools.
Overview
mediause mcp starts a stdio-based MCP server. It maps semantic CLI commands to discrete tools that LLM agents can discover and call automatically.
# Start MCP server locally mediause mcp # Limit to specific plugins mediause mcp --plugins weibo,twitter
Core Tools
The server provides built-in tools for session management:
- mediause_auth_login: Initialize browser login flow
- mediause_auth_list: List all stored credentials
- mediause_auth_health: Verify if a session is still valid
- mediause_use_account: Securely bind a browser context to an account
Site Plugin Tools
Each dynamic command is mapped to a tool using the site_capability_action naming convention.
# Example tool mappings: # mediause weibo search hot -> weibo_search_hot # mediause twitter post feed -> twitter_post_feed
How to call
Agents call these tools using standard MCP JSON-RPC. Arguments mirror the CLI flags (without the -- prefix).
// MCP tools/call example { "method": "tools/call", "params": { "name": "weibo_search_hot", "arguments": { "limit": 10 } } }
Return Structure
MediaUse tools return structured results rather than plain strings. This ensures that agents can reliably extract and use the data without fragile regex or text parsing.
// Structured Data Example { "status": "ok", "data": { "hot_topics": [ { "title": "...", "score": 99 } ] }, "task_id": "task_123" }
Agent configuration
Add MediaUse to your agent's config (e.g., Claude Desktop, Cursor, or custom MCP clients).
// Claude Desktop config example { "mcpServers": { "mediause": { "command": "mediause", "args": ["mcp"] } } }