Varrd
About
VARRD is the Cursor for trading — an AI-native quant research engine that turns domain knowledge into statistically validated trading edges. Describe any idea in plain English, and VARRD loads real market data, builds the pattern, and runs institutional-grade statistical tests to
Details
- Author
- augiemazza
- Downloads
- 368
- Categories
- Finance, Other, Search
Jump to
- Scans live strategies for actionable signals (free)
- Researches trading ideas via multi‑turn AI workflow (~20–30¢)
- Discovers edges autonomously (~20–30¢)
- Enforces statistical guardrails (K‑penalty, multiple testing correction)
- Supports futures (CME), US equities, and crypto (Binance)
- Provides exact entry, stop‑loss, and take‑profit prices
Setting up with Highlight
This MCP is not yet compatible with Highlight’s one-click setup. However, you can still use it with Highlight by following these steps:
- Download and install Highlight from highlightai.com/download
- Navigate to the plugins tab and select "Add Custom Plugin"
-
Configure the plugin with the settings below
Plugin Name
VarrdCommand (node, npx, python, etc.)Please refer to the README for specific instructions on how to obtain API keys or other required environment variables.
- Enable "Start Automatically" if you want the plugin to start when Highlight launches
From the repository
Install via pip install varrd, then use the Python library or CLI tool. For AI agents (Claude Desktop, Claude Code, Cursor), add the MCP server configuration with transport type streamable-http pointing to https://app.varrd.com/mcp. Typical workflow: varrd scan to see current signals, then varrd research "your idea" for a multi‑turn statistical test.
Claude Desktop / Cursor
Paste into your MCP client config file to install this server.
{
"mcpServers": {
"varrd": {
"augiemazza-varrd": {
"args": [],
"command": "varrd"
}
}
}
}McpServers
{
"augiemazza-varrd": {
"args": [],
"command": "varrd"
}
}VARRD — Trading Edge Discovery
You can ask any AI to backtest a trading strategy. It will happily do it. The results will look great. And they will be wrong. Not wrong like "off by a little." Wrong like "the edge never existed and you'll find out with real money." Quantitative testing is full of invisible landmines — dozens of statistical biases, penalties, and correctional procedures that determine whether a result is real or an artifact of how you tested it. Things like what must be penalized when you test multiple variations. What can and can't be tested on the same data. When a result that looks significant is actually meaningless. How to tell the difference between a strategy that works and a strategy that just happened to overlap with a bull market. Why looking at your out-of-sample results "just to check" permanently contaminates them. Professional quants at top firms spend years learning these rules. Most of them still get it wrong sometimes. When you ask an AI to "backtest this strategy," it skips all of it. Not maliciously — it just doesn't know what it doesn't know. And neither do you. That's the problem. VARRD is a quant research system built on the statistical framework that institutional firms use internally. Every bias accounted for. Every penalty applied. Every checkpoint enforced — not by documentation or best practices, but by the structure of the system itself. You literally cannot skip the steps that need to happen, because the workflow won't let you. You bring a messy idea in plain English. VARRD does the math and gives you a verdict: edge or no edge — with exact entry, stop-loss, and take-profit prices. If the edge is real, you'll know. If it's not, you'll know that too — and you found out for 25 cents instead of $25,000 in live losses. ``
pip install varrd
`
---
Why This Matters
There are things in quantitative testing that are near-invisible to the human eye. Not complex — invisible. The kind of stuff that a PhD statistician catches on instinct after 15 years, that a trader learns the hard way after blowing up twice, that a quant at Citadel takes for granted but never explains because it's just "how things are done."
Things like:
- Why testing 5 RSI thresholds and picking the best one isn't the same as testing 1
- Why a strategy that "beats the market" might actually be behind the market
- Why every formula tweak is a statistical test, whether you think of it that way or not
- Why your out-of-sample validation becomes worthless the moment you use it to make a decision
- Why the number of observations matters in ways that aren't obvious
- Why significance at one horizon says nothing about significance at another
These aren't advanced topics. They're table stakes. And if even one of them is handled wrong, the whole result is unreliable.
VARRD handles all of them. Automatically. Invisibly. You don't configure anything. You don't set penalty parameters. You don't choose which corrections to apply. The system knows what needs to happen at each stage of research and it does it — the same way a quant at a top firm would, except it never forgets a step and it never cuts corners because it's 4pm on a Friday.
You don't need to know what any of this means. That's the point.
---
Quick Start — Python
``python
from varrd import VARRD
v = VARRD() # auto-creates free accountSign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
