# strategy-lab

Stress-tests lowcap, perp and LP strategies over thousands of simulated paths and returns the whole distribution plus a seed that reproduces the run.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI
- Author: kellyruns
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/kellyruns/kellyruns
- Updated: 2026-08-14
- HTML: https://affpapa.org/en/nevibecode/mcp/strategy-lab

## What it does

You hand the server a strategy as parameters, not prose. It runs that strategy across many simulated trajectories and returns the full distribution instead of a single backtest figure, so the tail where losses pile up stays visible. Five tools ship with it: agent state, the certification thresholds, a lab summary, a filtered strategy list with certification rollups, and one strategy in full.

Connection goes over HTTP with no key and no auth, protocol `2025-06-18`. The endpoint is read-only on purpose. Nothing in it submits a strategy, pays for a run or moves funds, since an unauthenticated address is the wrong place for that.

The PRNG seed is the part that matters. Feed it back and the run repeats exactly, which lets you argue with a verdict by rerunning it. Execution is still paper mode, so fills are simulated and the EVM path sits wired but unarmed. A certification describes a distribution. It does not forecast returns, and none of this is financial advice.

## Requirements
- Remote server: `https://kellyruns.xyz/mcp`

## Install

### Claude Code (native platform)

**Platform CLI: remote server**
```bash
claude mcp add --transport http strategy-lab https://kellyruns.xyz/mcp
```

### Codex

**Platform CLI: remote server**
```bash
codex mcp add strategy-lab --url https://kellyruns.xyz/mcp
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "strategy-lab": {
            "url": "https://kellyruns.xyz/mcp"
        }
    }
}
```

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "strategy-lab": {
            "type": "http",
            "url": "https://kellyruns.xyz/mcp"
        }
    }
}
```

### Gemini CLI

**Platform CLI: remote server**
```bash
gemini mcp add -t http strategy-lab https://kellyruns.xyz/mcp
```

## When to use it
- run a perp strategy across thousands of paths and inspect the tail
- reproduce a past run exactly from its saved seed
- check whether a strategy clears the certification thresholds
- rank several strategies by median PnL and worst case
- see how many more runs the testing pool can fund

## Example prompts
- Pull the certification thresholds with `get_certification_gate` and tell me what clearing them actually means.
- List LP strategies with status passed and show their certification rollup.
- Give me one strategy in full with the seed from its last run so I can repeat it.

## Similar
- [strategy-pivot-designer](https://affpapa.org/en/nevibecode/skills/strategy-pivot-designer.md) — skill · 3300 installs
- [strategy-compare](https://affpapa.org/en/nevibecode/skills/strategy-compare.md) — skill · 2300 installs
- [strategy-framework](https://affpapa.org/en/nevibecode/skills/strategy-framework.md) — skill · 501 installs
- [prediction-market-strategy](https://affpapa.org/en/nevibecode/skills/prediction-market-strategy.md) — skill · 313 installs
- [Monte-Neo strategy verifier](https://affpapa.org/en/nevibecode/mcp/monte-neo-strategy-verifier.md) — MCP server · ★ 8
- [llmquant-portfolio-lab](https://affpapa.org/en/nevibecode/skills/llmquant-portfolio-lab.md) — skill · 479 installs

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md