Forge strategies. Trust only proof.
Forven is a free, open-source, self-hosted workspace where AI research agents, a backtesting gauntlet, and a risk command center turn fragile ideas into evidence — local-first, your keys, no account.
- Local-first
- BYO keys
- No account
- AGPL-3.0
Strategies, data, and runs stay on your machine.
Your exchange and LLM keys never leave your machine.
Clone and run — no sign-up, no license key.
Free and open source. Read every line.
A research process with hard gates.
Forven is not a signal feed. It is a local process for turning fragile ideas into candidates that have earned further attention.
Build the cohort
Build strategies in the Strategy Creator — from indicators, plain English, or code — or let agents and your own AI client draft them, all in one local lab.
Run the gauntlet
Push each idea through screen, walk-forward gauntlet, candidate, and live-review stages with explicit gates.
Promote with evidence
Use out-of-sample metrics, audit trails, and paper-trade behavior to decide what survives and what gets retired.
Most strategies fail. Forven makes failure useful.
Each stage asks a harsher question of the same idea. If the out-of-sample evidence stops holding up, the strategy stops moving forward.
Fast filters remove fragile ideas before deeper testing.
Walk-forward windows test survival outside the fit period.
Paper trading and review catch behavior that backtests hide.
Only a small set earns live-routing consideration.
The point is not optimism.
The point is disciplined rejection. Forven helps you learn why an edge died before that lesson costs live capital.
A risk command center that can halt itself.
Drawdown, daily-loss, and portfolio-risk limits run as live gauges with circuit breakers. Cross a threshold and Forven halts new live entries — or closes positions outright — before you do.
Breakers gate the price feed, trade routing, and account before a bad fill compounds.
One confirmed control closes every open live position at market and halts new entries while you investigate.
Inspect the lab before you trust the output.
Forven keeps the workflow visible: strategy cohorts, current stages, recent out-of-sample behavior, and the trail of decisions that moved an idea forward.
| ID | Strategy | Stage | Trades | OOS | Status |
|---|---|---|---|---|---|
| S00241 | rsi_mean_reversion_btc | live | 127 | +2.4% | pass |
| S00512 | vwap_reclaim_eth | candidate | 38 | +0.9% | pass |
| S00693 | volume_anomaly_sol | gauntlet | 212 | -1.1% | fail |
| S00718 | breakout_confirm_btc | gauntlet | 94 | +0.4% | watch |
- Local by defaultStrategies, data, and runs stay on your machine unless you choose to connect external services.
- Bring your own keysUse your own exchange and LLM keys. Forven does not need to resell model tokens to be useful.
- Auditable gatesEvery promotion has a stage, a metric trail, and a reason. No hidden signal feed, no black-box handwave.
Prove it forward before any live capital.
Survivors trade forward on live market data through the same engine that ran their backtest, with every fill simulated on your machine — so behavior that backtests hide shows up where it’s safe.
Strategies trade forward on the live Binance feed — the same series their backtests ran on.
One execution kernel runs entries, exits, stops, sizing, fees, and funding in backtest, paper, and live.
Every fill is simulated in-app with modeled costs. No order is sent to any exchange.
Sample session from an earlier build, for illustration — strategy names, charts, and PnL are not real or projected returns.
Co-researchers, not co-pilots.
Agents are optional co-researchers included for everyone. They can draft hypotheses, inspect failed runs, and explain drawdowns, but the gauntlet still decides what earns attention.
They run on your own model keys — only the providers you connect and the models you enable, never resold. The lab, the research loop, and the agent layer are all free and open source.
- Drafts hypothesesand develops the most promising from a daily research budget
- Writes strategy coderegistered through the same safety guard as yours
- Inspects gauntlet resultsreasons about why edges died
- Writes post-mortemson failed candidates and live drawdowns
| Agent | Role |
|---|---|
| Brain | Orchestrates: routes strategies, enforces gates |
| Strategy Developer | Designs, codes, and registers strategies |
| Simulation Agent | Backtests, optimization, the gauntlet |
| Risk Manager | Reviews promotions; watches paper and live |
| Quant Researcher | Hypotheses, data checks, post-mortems |
| Full-Stack Engineer | Operator-triggered triage and repair |
Or bring the AI you already use.
Forven ships a Model Context Protocol server. Connect your AI client and it can design, backtest, and validate strategies in your lab — through the same gauntlet and the same gates as everything else.
A genuine rejection is the system working: the client reads the gate report and fixes what failed. It can't skip a gate, and promotion to live always stays with you.
Connect an AI client →Ready-made configs for Claude, Codex, Cursor, ChatGPT Desktop, VS Code, and more are in the app under Integrations → AI clients.
- Reads the workspacecontext, the strategy template, datasets, and the gotchas
- Writes and registers strategiesthrough the same safety guard and look-ahead probe
- Runs the evidencebacktests, optimization, and the persisted robustness suite
- Reads the gate reportand promotes only through the real gates — live stays yours
Free and open source.
Every Forven feature is free and open source. Clone the repository and run one launcher script — the whole stack runs on your own machine. Bring your own LLM key when you want agents to help draft, debug, and review.
Then run start_all.ps1 on Windows, or follow the install guide's short setup for macOS and Linux.
- Clone the repo and run one launcher script — no installer, no account.
- Bring your own exchange and LLM keys; nothing is resold or proxied.
- Runs on Windows, macOS, and Linux from source.