Blave-TW

Blave Agent — DevOps skill for Claude Code

DevOps community

Quant infrastructure for AI agents — a free, open-source macOS workspace where your Claude Code or Codex turns a trading idea into a backtested strategy and runs it live.

How to install Blave Agent

This entry records only its repository, not the path inside it, so there is no exact command to give. Open Blave-TW/blave-agent and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Blave Agent does

Quant infrastructure for AI agents — a free, open-source macOS workspace where your Claude Code or Codex turns a trading idea into a backtested strategy and runs it live.

Alternatives in DevOps

Claudable — Claudable is an open-source web builder that leverages local CLI agents, such as Claude Code, Codex, Gemini CL 4k ★
Skill Claw — OpenClaw instance administration — manage hosts across macOS, Ubuntu/Debian, Docker, OCI, and Proxmox 2.8k ★
Second Brain Cloudflare — One memory layer, every AI tool 734 ★

README

Blave Agent

**Agentic Quant Workspace**

Turn Your Agent into a Quant

Free and open source. Connect your Claude Code or Codex. You describe the idea; it writes the strategy, runs the backtest, and trades it live.

**English** | [繁體中文](README.zh-TW.md)

![License: Apache-2.0](https://img.shields.io/badge/license-Apache--2.0-lightgrey) ![Platform: macOS](https://img.shields.io/badge/platform-macOS-lightgrey)

https://github.com/user-attachments/assets/7b33edb7-9c65-4e19-854a-40295c6e8b74

[Download the macOS app](https://github.com/Blave-TW/blave-agent/releases/latest) · [Quick start (from source)](#quick-start) · [Run it with your computer off](https://blave.org/agent/en)

Star the repo if this is useful — and Watch › Releases to get notified of new versions.

What Makes It Different

Backtests That Check Whether It Was Luck

Every Type A backtest runs a Monte Carlo permutation test by default (MCPT, lib/validation.py) and records a p-value: could shuffled data have done as well?
A parameter scan (lib/param_scan.py) looks for a plateau of parameters that all work, not the single best cell.
Rolling walk-forward (lib/walk_forward.py) measures out-of-sample performance.
The fee has to match the real market. A fee of 0 is flagged by lib/quality_check.py and treated as a bug.
One idea gets one backtest by default. A poor result is reported as it is; the agent does not quietly re-tune the parameters until the numbers look good (see Iteration Brakes in `AGENTS.md`).

See Whether Live Runs the Code You Backtested

A backtest pins a version of the strategy. If the code running live no longer matches that version, the strategy is flagged — the web workspace shows "Live · file changed" instead of a clean "Live". The flag does not stop the strategy from running. It applies