AlphaForge
The quant CLI your AI agent drives — backtest to Pine v6
12 followers
The quant CLI your AI agent drives — backtest to Pine v6
12 followers
AlphaForge is a local-first quant CLI built for AI agents. Ask Claude Code (or any MCP client) to explore strategies overnight: backtest → Optuna TPE optimization → walk-forward validation → TradingView Pine Script v6 export, all from one standalone binary (macOS/Windows). Free Trial included — no signup. The visualizer, order-execution webhook, and MCP server are Apache-2.0 OSS and hit 1.0 together.
finally something that doesn't make me babysit a python env just to run a backtest, the pine export is a really nice touch honestly
@lander_feng Thank you — the Python env thing is the entire reason it ships as a single binary. No venv, no dependency resolution, no "works on my machine" step standing between you and a number. If you want to see it end to end, alpha-forge demo runs the whole loop on bundled synthetic data with no network and no signup.
One fair warning on Pine so it doesn't surprise you: export is the one thing the free Trial blocks. Data up to 2023-12-31 and 50 optimization trials are open, but pine generate is paid-tier. pine import works either way, so you can pull an existing script in and see how it maps onto the strategy JSON.
Curious what you'd want to export to — are you running TradingView alerts for execution, or is it charting only?
Love that the visualizer, webhook, and MCP server are Apache-2.0 and ship to 1.0 together, that kind of commitment to keeping the core tooling open while monetizing the binary is rare and really respectful to the community.
@orlando_gui Thank you — that split wasn't a marketing decision, it was the only version I could defend. The three open pieces are the ones that touch things I have no business asking you to trust blindly: your broker connection, your data, and whatever your agent is doing on your machine. Nobody should wire a webhook to a live account whose source they can't read.
The closed part is the backtest and optimization engine — that's the piece I sell, and I'd rather say that plainly than muddy it. They're all at github.com/alforge-labs if you want to poke around; the 1.0 tags went out together on purpose so nobody ends up on a version mismatch between the core and the tooling.