bindai-mcp

# BindAI MCP `bindai-mcp` provides MCP integration functionality for connecting BindAI applications and agents to external tool services. The package exposes a small public API centered around `MCPClient` and `MCPTool`. > **Current status:** The package currently provides a lightweight HTTP-based tool discovery and invocation bridge. Full Model Context Protocol interoperability is part of the BindAI MCP roadmap. ## Features * MCP integration foundation * Remote tool discovery * Representation of discovered tools as BindAI tools * Integration of external tools into BindAI applications * Lightweight public Python API * Separation between MCP integration and BindAI core functionality ## Installation `bindai-mcp` is part of the BindAI workspace. From the BindAI repository root: ```bash uv sync ``` The package can then be used through the workspace environment. For public package installation, install the published `bindai-mcp` package from PyPI when the corresponding release is available. ## Basic Usage The public API can be imported as: ```python from bindai_mcp import MCPClient, MCPTool ``` `MCPClient` provides the client-side integration, while `MCPTool` represents a remotely exposed tool as a BindAI tool. The current implementation uses an HTTP service exposing tool discovery and invocation endpoints. A typical integration flow is: ```text BindAI Agent / Application │ ▼ MCPClient │ ▼ HTTP Tool Service │ ┌────┴────┐ ▼ ▼ Tool A Tool B ``` The current client expects: ```text GET /tools POST /call ``` The exact request and response behavior is provided by the package implementation. ## Public API ### MCPClient `MCPClient` is the primary client interface for discovering external tools. ```python from bindai_mcp import MCPClient ``` Example: ```python client = MCPClient("…