Forge

You run an MCP server that exposes a formal, adapter-like interface to interact with an OpenAPI-described API. This server makes it easier to connect multiple agents to the 1Forge API via the Multi-Agent Conversation Protocol, enabling coordinated data retrieval and actions across clients.

How to use

Start the MCP server locally and connect an MCP client to it. You will interact with the server through its standard input/output interface or through a defined transport, and you can configure it to load a JSON configuration that governs security and runtime behavior. Your MCP client can then request data, trigger actions, and receive responses in a structured, channel-based flow that supports multiple agents collaborating on tasks.

In practice, you typically run the server in a local environment and point your client to the server’s endpoint. You’ll configure security parameters and the path to the server’s configuration, then start exchanging messages following the MCP conventions. The general pattern is to initialize the server, authenticate if needed, load the configuration, and then begin sending requests that the server forwards to the underlying API.

How to install

Prerequisites you need before installation are Python 3.9 or newer, plus pip and uv. Ensure these are available on your system.

Step 1: Clone the repository and navigate into the project folder.

git clone <repository-url>
 cd mcp-server

Step 2: Install dependencies. Use the development extras to include all tooling for linting, type checking, and tests.

pip install -e ".[dev]"

If you prefer to use uv for development, you can install the editable package via uv as well.

uv pip install --editable ".[dev]"

Step 3: Prepare to run the server. You can load configuration from a file path or pass a JSON string with the configuration.

Step 4: Start the server using the stdio transport as shown in the execution example.

Configuration and running notes

The server accepts runtime configuration via environment variables. You can set a path to a JSON configuration file with CONFIG_PATH, or provide a JSON string via CONFIG. You can also supply SECURITY-related environment parameters for access control.

To start the server in stdio mode from the command line, run: python mcp_server/main.py stdio.

Example environment variables you might use include the following (placeholders shown):

CONFIG_PATH=/path/to/mcp_config.json
CONFIG='{"transport":"stdio","server":"forge_mcp"}'
SECURITY='{"apiKey":"YOUR_API_KEY"}'

Available tools

lint

Run linting to detect code quality issues and ensure consistent style.

format

Format the codebase to adhere to style guidelines.

static-analysis

Run static analysis tools for type checks and security reviews.

test

Execute the test suite to verify functionality and generate coverage reports.