Schwab Model Context Protocol Server

The Schwab Model Context Protocol (MCP) Server connects your Schwab account to LLM-based applications (like Claude Desktop or other MCP clients), allowing them to retrieve market data, check account status, and (optionally) place orders under your supervision.

Features

• Market Data: Real-time quotes, price history, option chains, and market movers.
• Account Management: View balances, positions, and transactions.
• Trading: comprehensive support for equities and options, including complex strategies (OCO, Bracket).
• Safety First: Critical actions (like trading) are gated behind a Discord approval workflow by default.
• LLM Integration: Designed specifically for Agentic AI workflows.

Quick Start

Prerequisites

• Python 3.10 or higher
• uv (recommended) or pip
• A Schwab Developer App Key and Secret (from the Schwab Developer Portal)

Installation

For most users, installing via uv tool or pip is easiest:

# Using uv (recommended for isolation)
uv tool install git+https://github.com/jkoelker/schwab-mcp.git

# Using pip
pip install git+https://github.com/jkoelker/schwab-mcp.git

Authentication

Before running the server, you must authenticate with Schwab to generate a token file.

# If installed via uv tool
schwab-mcp auth --client-id YOUR_KEY --client-secret YOUR_SECRET --callback-url https://127.0.0.1:8182

# If running from source
uv run schwab-mcp auth --client-id YOUR_KEY --client-secret YOUR_SECRET --callback-url https://127.0.0.1:8182

This will open a browser window for you to log in to Schwab. Once complete, a token will be saved to ~/.local/share/schwab-mcp/token.yaml.

Running the Server

Start the MCP server to expose the tools to your MCP client.

# Basic Read-Only Mode (Safest)
schwab-mcp server --client-id YOUR_KEY --client-secret YOUR_SECRET

# Streamable HTTP (for reverse proxies / remote MCP connectors)
schwab-mcp server --client-id YOUR_KEY --client-secret YOUR_SECRET \
  --http --host 127.0.0.1 --port 8000

# With Trading Enabled (Requires Discord Approval)
schwab-mcp server \
  --client-id YOUR_KEY \
  --client-secret YOUR_SECRET \
  --discord-token BOT_TOKEN \
  --discord-channel-id CHANNEL_ID \
  --discord-approver YOUR_USER_ID

# Or approve over Signal instead (requires a local bbernhard/signal-cli-rest-api
# daemon running in MODE=json-rpc or json-rpc-native; other modes cannot
# stream replies and silently consume them)
schwab-mcp server \
  --client-id YOUR_KEY \
  --client-secret YOUR_SECRET \
  --signal-account +15555550100 \
  --signal-approver +15555550199

Default transport is stdio (Claude Desktop and most local MCP clients). Use --http for FastMCP streamable-http when fronting the server with a gateway or remote connector (MCP endpoint is /mcp on the bound host:port).

Note: For trading capabilities, you must set up an approval backend: either a Discord bot (see Discord Setup Guide) or Signal via a local signal-cli REST daemon running in MODE=json-rpc (or json-rpc-native). Configure either backend, not both. (The signal extra declares the websockets dependency explicitly; a default install already includes it transitively.)

Configuration

You can configure the server using CLI flags or Environment Variables.

Container Usage

A Docker/Podman image is available at ghcr.io/jkoelker/schwab-mcp.

podman run --rm -it \
  --env SCHWAB_CLIENT_ID=... \
  --env SCHWAB_CLIENT_SECRET=... \
  -v ~/.local/share/schwab-mcp:/schwab-mcp \
  ghcr.io/jkoelker/schwab-mcp:latest server --token-path /schwab-mcp/token.yaml

Available Tools

The server provides a rich set of tools for LLMs.

📊 Market Data

💼 Account Info

💸 Trading (Requires Approval)

Placing an order is two steps: preview it, then place it by ID. Every preview_* tool builds the order and calls Schwab's preview API to return the projected order details plus a preview_id; place_previewed_order then submits that exact previewed order — no re-derivation from parameters, so the LLM can't hallucinate a different order than what was reviewed. Replacing an order uses the same safe flow: preview a complete single-leg equity, option, or equity trailing-stop description with preview_replacement_order, then call replace_previewed_order for the bound target order.

(See full tool list in src/schwab_mcp/tools/)

Development

To contribute to this project:

# Clone and install dependencies
git clone https://github.com/jkoelker/schwab-mcp.git
cd schwab-mcp
uv sync

# Run tests
uv run pytest

# Format and Lint
uv run ruff format . && uv run ruff check .

Mutation testing

Run a local mutation campaign with make mutation (two workers). The weekly GitHub Actions workflow runs Sundays at 07:29 UTC and also supports manual runs. CI limits the mutation step to 20 minutes and saves reports for 14 days; surviving mutants are informational and do not fail the build by score.

License

MIT License.