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:

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:

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:

BindAI Agent / Application
           │
           ▼
       MCPClient
           │
           ▼
     HTTP Tool Service
           │
      ┌────┴────┐
      ▼         ▼
   Tool A     Tool B

The current client expects:

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.

from bindai_mcp import MCPClient

Example:

client = MCPClient("http://localhost:8000")
tools = await client.list_tools()

MCPTool

MCPTool represents a remotely available tool using BindAI's tool abstraction.

from bindai_mcp import MCPTool

Discovered tools can be represented as BindAI tools and integrated with BindAI agent execution.

BindAI Integration

The MCP integration keeps external tool connectivity separate from BindAI's core execution and agent abstractions.

Conceptually:

                 BindAI
                    │
           ┌────────┴────────┐
           │                 │
         Agent          Application
           │                 │
           └────────┬────────┘
                    │
                MCPClient
                    │
                    ▼
             External Tool
               Service
                    │
              ┌─────┴─────┐
              ▼           ▼
           MCPTool     MCPTool

This separation allows the MCP integration layer to evolve independently from the BindAI core framework.

MCP Roadmap

The BindAI MCP package is intended to evolve toward full MCP protocol support.

Planned capabilities include:

• Standard MCP client communication
• MCP tool discovery
• MCP tool invocation
• MCP resource support where appropriate
• BindAI agent integration
• Authentication and configuration
• MCP server integration where appropriate
• Documentation and examples
• Protocol-level interoperability testing

The current HTTP tool bridge should therefore be considered the foundation for the broader MCP integration rather than a complete MCP protocol implementation.

Development

From the repository root:

uv sync

Run the complete test suite:

uv run pytest

Run type checking for the package:

uv run mypy packages/bindai-mcp

Build the package:

uv build --package bindai-mcp

Package Structure

The package follows the standard BindAI package layout:

bindai-mcp/
├── pyproject.toml
├── README.md
├── src/
│   └── bindai_mcp/
│       ├── __init__.py
│       └── client.py
└── tests/
    └── test_client.py

API Stability

The package currently exposes:

from bindai_mcp import MCPClient, MCPTool

These are the intended public entry points of the package.

Internal implementation details should be treated as subject to change unless explicitly documented as public API.

Documentation

Full BindAI documentation is available at:

https://docs.bindai.dev

License

BindAI is released under the MIT License.