Back to Chat
Model Context Protocol (MCP)

Connect AI-Mate to Your AI Agents.

Seamlessly integrate AI-Mate's guided thinking engine, knowledge vault, and multi-tier AI models with Claude Desktop, Cursor, Antigravity, OpenCode, and custom AI agents.

Live MCP Endpoint

Use this standard HTTP/SSE endpoint in any MCP-compatible AI client:

https://ai-mate.app/mcp

⚡ One-Click AI Client Configurations

Select your favorite AI agent or IDE environment to get the ready-to-paste setup:

claude_desktop_config.json
{
  "mcpServers": {
    "ai-mate": {
      "url": "https://ai-mate.app/mcp"
    }
  }
}

🛠️ Available MCP Tools

AI-Mate exposes the following standardized functions to connected agents:

Tool NameCategoryCapability & Description
start_guided_thoughtAction / ThinkingStarts a guided thinking session from a raw user thought, asking 1 clarifying question first.
continue_guided_thoughtAction / ThinkingProgresses user responses through the thinking path, suggesting action chips when ready.
create_simple_resultGenerationProduces one structured output (clarity_summary, idea_directions, option_comparison, simple_plan).
export_simple_planExportExports a simple plan into ready prompts for Antigravity, Claude, Codex, Hermes, OpenCode.
save_resultStorageSaves a generated thought result directly into the user's permanent AI-Mate vault.
capture_artifactKnowledge OSProgrammatically captures a structured decision, insight, hypothesis, or summary artifact.
search_artifactsKnowledge OSSearches through user's captured knowledge artifacts with keyword and type filters.
list_available_modelsCatalogLists all available multi-tier models (DeepSeek V4 Flash, Hy4 770B MoE, GLM 5.3) & token rates.
get_wallet_balanceAccountChecks authenticated user's current pay-as-you-go credit balance and selected model.
Open Standards: Built with official Model Context Protocol (MCP) specification supporting streamable HTTP & SSE.
Secure & Private: User knowledge and chat history remain strictly isolated under OAuth and scoped security controls.