taas-mcp
A local MCP server that acts as a thin proxy to expose Tread's trading data and order tools to a local LLM, fetching tool catalog and forwarding calls to the Tread backend.
README
taas-mcp
A local MCP server that exposes Tread's trading-data and order tools to a local LLM. It is a thin, generic proxy: it holds no hardcoded tools — on startup it fetches the tool catalog from your Tread backend and forwards each call there. All reasoning happens in your local agent; Tread executes the individual tool calls.
Backend dependency. This client talks to the Tread backend
mcp_apiendpoints (GET /api/mcp/tools,POST /api/mcp/dispatch,GET /api/mcp/guide), which must be deployed and theMCPfeature flag enabled for your user. Design / PRD: Linear project "MCP Server — Local-LLM Trading Data & Orders".
What it exposes
Whatever the backend registry advertises — market data, balances, portfolio analytics, order status, and trading (single, market-maker, delta-neutral, and batch orders, plus cancel / pause / resume / set leverage). Because the catalog is fetched live, adding or changing a tool on the backend needs no new release here.
It also exposes the authoritative operating guide as the
tread_operating_guide prompt, fetched live from the backend — load it before
driving the tools.
Install
First install uv — this is the only prerequisite.
The uvx command used below ships with uv (it's shorthand for uv tool run),
so there is nothing else to install:
# macOS / Linux
curl -LsSf https://astral.sh/uv/install.sh | sh
# or, with Homebrew
brew install uv
Then install and run straight from the repo — no clone required. uvx fetches
the package from GitHub, builds it in an isolated environment (provisioning a
compatible Python automatically), and runs the taas-mcp entry point:
uvx --from git+https://github.com/tread-labs-public/taas_mcp.git taas-mcp
The server speaks MCP over stdio, so it's normally launched by your MCP client (see Connect your MCP client) rather than run interactively. The command above is the same one you put in the client config.
Pin to a branch or tag by appending @<ref>:
uvx --from git+https://github.com/tread-labs-public/taas_mcp.git@main taas-mcp
To install it once as a persistent tool on your PATH (so the command is just
taas-mcp), use uv tool instead:
uv tool install git+https://github.com/tread-labs-public/taas_mcp.git
uv tool upgrade taas-mcp # pull a newer version later
[!NOTE]
uvxcaches the build. After the backend ships changes that only affect this client, force a rebuild withuvx --refresh --from git+https://github.com/tread-labs-public/taas_mcp.git taas-mcp.
Configuration
| Var | Required | Description |
|---|---|---|
TREAD_URL |
yes | Base URL of the Tread server, e.g. https://app.tread.fi |
TREAD_MCP_TOKEN |
yes | Your MCP token (generate/regenerate it in the Tread web UI) |
TREAD_READONLY |
no | 1 to hide all write/trading tools |
TREAD_TIMEOUT |
no | HTTP timeout in seconds (default 30) |
The token authenticates as you; every call is scoped to your account
server-side. One running server == one Tread user. The token expires after a
peiod (default 30 days) of inactivity (it is refreshed on each use); once expired, requests
return 401 and you regenerate it from the Tread web UI.
Connect your MCP client
Point your MCP client at the same uvx command. The client launches the server
on demand and passes configuration through env.
Claude Code (CLI)
claude mcp add tread \
--env TREAD_URL=https://app.tread.fi \
--env TREAD_MCP_TOKEN=your-token-here \
-- uvx --from git+https://github.com/tread-labs-public/taas_mcp.git taas-mcp
Add --env TREAD_READONLY=1 for a read-only setup.
Claude Desktop / generic MCP client (JSON)
{
"mcpServers": {
"tread": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/tread-labs-public/taas_mcp.git",
"taas-mcp"
],
"env": {
"TREAD_URL": "https://app.tread.fi",
"TREAD_MCP_TOKEN": "your-token-here"
}
}
}
}
For a read-only setup, add "TREAD_READONLY": "1" to env.
[!NOTE] The client must be able to find
uvxon itsPATH. If it can't (common on macOS GUI apps), use the absolute path fromwhich uvxascommand, e.g./opt/homebrew/bin/uvx.
Develop
Clone the repo for local development, then sync and test:
git clone https://github.com/tread-labs-public/taas_mcp.git
cd taas_mcp
uv sync
uv run pytest
Run your local checkout directly (point the client at this command to test changes before pushing):
uv run --directory /path/to/taas_mcp taas-mcp
Safety
The local agent owns risk. Tread runs no server-side agent for these calls —
there is no automated margin management, invariant enforcement, or liveness
rescue behind them. See the tread_operating_guide prompt for the operating
contract.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.