AlgoBreeder
WorkshopWorkflowMarketsAgentsPricingDocs
Open AppResearcherStart free

Claude · Cursor · Codex · your trading lab

AlgoBreeder

Connect the AI you already use — Claude, Cursor, Codex — and give it every tool it needs to draft, train, backtest, and stress-test a trading strategy fast. No in-app chatbot. Your agent runs a real research lab. Paper first.

Connect your agent
Researcher plans

› Hook up Claude or Cursor · your agent drafts, trains, and backtests for you

See it in action↓Interactive labs · what your agent can do

Interactive workshop

Feel what your agent can do — before you wire it up

Four educational labs: draft flow, overfit traps, Monte Carlo stress, and a dashboard preview. Same loop your LLM will run once connected — create, train, backtest, pressure-test. Not live jobs. Paper-side research.

Agent stream preview. Rerun when idle.

Educational previewWhat · Why · How
What
Scripted agent/MCP stream: Community holdout + trade-shuffle MC, then a walk-forward idea.
Why
See the validation order agents run locally, not a live job, lease, or in-app chatbot.
How
Autoplays when this lab scrolls into view; Rerun replays. Metrics are fiction. Real rolling WF + Bayesian = Researcher.
educational preview · agent/MCP stream
ready

awaiting stream…

telemetry

awaiting stream…

MCP · playbook · educational preview

› streaming · agent tools appear in order

No chatbot · MCP + playbook

Researcher when ready →
MCP docs

Why Researcher exists

Fast AI drafts are easy. Knowing the edge survives is the product.

Your LLM can invent rules in seconds. AlgoBreeder is the lab that makes it prove them — holdout, stress tests, and deeper Researcher checks — before you trust size. Community starts free; Researcher finishes the hard parts.

Researcher plansSee every stage →

Community

Holdout training, trade-shuffle Monte Carlo, sensitivity, anchored walk-forward.

Researcher

Rolling walk-forward + Bayesian/adaptive search: the folds Community cannot run.

What your agent runs

From idea to stress-tested draft — one loop your LLM can drive

Free path: draft a strategy → train → backtest → Monte Carlo → paper sims. Researcher adds the harder out-of-sample depth, insights, multi-asset training, and educational exports for TradingView / MT5 / cTrader. We do not encourage live trading.

Active stage · 01

Community path

Strategy

Versioned rules with entry, exit, and sizing — start from Dual EMA trend or MA trend follower, or draft new ones via MCP agents.

Included on Community

Your agent helps here via MCP: run_training_session / playbook. Serious rolling depth, insights, and broker exports need Researcher.

Rolling + Bayesian

Walk-forward that Community cannot run

Insights & advanced MC

Regimes · overfit heat · return perturbation

TV · MT5 · cTrader

Exports that run on those platforms · paper first

Agent step · get_research_playbook · gated tools appear when Researcher is unlocked

After the gears · market dials

Real pack classes: retrieve every major, not a three-symbol teaser

Community packs include G7 FX majors plus liquid crosses, metals, crypto, Brent, and continuous futures. Chips are for browsing; your agent pulls any pack symbol with ensure_community_data / download_market_data / inspect_instrument. Community depth ~365d; Researcher targets ~5y on 1h/4h and ~10y daily for FX.

Forex majors

Full Dukascopy FX major set in the community pack; dials below are samples. Agents retrieve any pack major via ensure_community_data / download_market_data.

All 10 majors above · retrieve via MCP

Demo · EURUSD · G7 majors + liquid crosses · ~365d community

Researcher · ~5y (1h) / ~10y (daily) FX

Metals · Crypto · Brent

XAU/XAG, BTC/ETH, and BRENT on the same pack class, not equities spot. Pull any listed symbol the same way.

Full metals/crypto/Brent pack class

Demo · XAUUSD · Metals · crypto · Brent · community depth

Researcher · multi-year windows within caps

Futures

Continuous contracts ES, NQ, CL, GC with session/cost models, not a full exchange roll-chain product. Agents can ensure each symbol from the pack.

All continuous futures in pack

Demo · ES · Continuous futures · session/cost model

Researcher · deeper futures history within caps

Agent step · list_community_datasets · ensure_community_data for any pack major

Open core · then agents

Flip the layer: see what stays closed

Free core gives your LLM a local lab it can call. Agents draft and validate through real tools — not a chatbot. Researcher keeps the deepest algorithms closed.

Apache 2.0 · strategy SDK, local MCP tools, backtesting, Training Center shell.

Star on GitHub
Architecture docs

Your LLM · a full trading lab

Arm Claude with everything it needs to build a strategy

Hook up Claude, Cursor, Codex, or VS Code. Your agent gets real tools to draft rules, train on history, backtest, and pressure-test ideas — so you go from chat to a researched strategy in one loop. Educational lab, paper first. No chatbot inside AlgoBreeder.

Works with

  • Cursor
  • Claude Desktop
  • VS Code
  • Codex
  • Claude Code
Connect your agent

agent toolbox

01 / 09

Read the research playbook

Community · free

Your agent pulls the recommended workflow text so it knows how to draft, validate, and stress-test a strategy draft in order.

When · First call after you arm the agent. Sets the sequence before any backtest.

Available on Community. Start with get_research_playbook.

Connect Claude · draft → train → backtest → stress-test · Plus tools need Researcher

The stack you already pay for

Traders spend more on tools that never ask if the edge is real.

Charts show price. Journals show history. Neither pressure-tests whether a system survives walk-forward folds, random paths, and regime shifts. Researcher is built for that question, at less than a TradingView Premium bill when you pay yearly.

ToolTypical spendWhat you getEdge?

  • TradingView Premium

    Charts, alerts, shallow backtests

    ~$720/yr

    Charts, alerts, shallow backtests

    No

  • Trade journals

    Log and review the P&L you already took

    ~$360–720/yr

    Log and review the P&L you already took

    No

  • Order-flow / AI charting

    Better eyes on the tape, still not robustness science

    ~$600–1,200/yr

    Better eyes on the tape, still not robustness science

    No

  • AlgoBreeder Researcher

    Rolling WF · Bayesian · MC · overfit · regimes · exports

    €49/mo

    Rolling WF · Bayesian · MC · overfit · regimes · exports

    Yes

Hundreds a year for charts and journals that never validate a system, or €49/mo for the layer that finds and kills bad systems before they cost you. Community proves the loop for free. Researcher is the no-brainer when you are building a real book.

Get Researcher yearly →

€590/yr · ≈ €49/mo · or jump to plans

Pricing

Free locally. Researcher for the hosted lab.

Community is the open-source local loop at €0. Researcher is €49/mo billed yearly for hosted Lab + MCP, Researcher+ validation (rolling WF, Bayesian search, insights, multi-asset, exports), and 2000 compute tokens / month. Paper first; live trading is not encouraged.

Community

Free

Open-source local stack on GitHub. Draft, train, backtest, and stress-test on your machine — no cloud account. Paper sims included; live trading is not encouraged.

  • ▹Your AI agent + local research / MCP
  • ▹Draft → train → backtest → stress-test
  • ▹Monte Carlo, sensitivity & core reports
  • ▹Starter strategies: Dual EMA · MA trend
  • ▹Full pack majors (~365d) · soft daily caps
Read the docs

Serious research depth

Billed yearly

Researcher

€590/yr

≈ €49/mo effective · less than TradingView Premium

Hosted AlgoBreeder Lab + MCP with Researcher+ algorithms: rolling walk-forward, Bayesian search, premium insights, multi-asset training, and educational TV / MT5 / cTrader exports. 2000 compute tokens / month. Paper first; we do not encourage live trading.

  • ▹Hosted Lab + hosted MCP (cloud account)
  • ▹Rolling walk-forward + Bayesian / adaptive search
  • ▹Premium insights · regimes · overfit heat · readiness
  • ▹Multi-asset training + stronger stress tools
  • ▹Educational exports: TradingView · MT5 · cTrader
  • ▹Starter strategies: Dual EMA trend · MA trend follower
  • ▹2000 compute tokens / mo · ~5y (1h) / ~10y (daily) FX
Open Researcher checkout

Checkout · unlock hosted Lab + Researcher+

Educational software · not financial advice

Same agent path · get_research_playbook · gated tools appear when unlocked

› Checkout opens Stripe in a new flow. Trader / Professional seats use the same path when Price IDs are set; this page never claims a gate cleared.

After checkout

Straight answers

Pricing closed the sell. FAQ cools the pulse: Community locally; Researcher via premium package or Stripe. Agents still use MCP + playbook, not a chatbot.

› Still deciding? Ask your agent for get_research_playbook. Researcher adds rolling WF, Bayesian, premium insights.

MCP docs →Researcher plans →

Checkout lives in the header · or jump to Plans

No. AlgoBreeder provides software for strategy research and validation. It does not recommend trades or guarantee outcomes.

Community is the free local GitHub stack (holdout, Monte Carlo, sensitivity, anchored walk-forward, paper sims, local MCP, ~365d packs, Dual EMA / MA trend starters). Researcher unlocks the hosted Lab + hosted MCP with Researcher+ algorithms: rolling walk-forward, Bayesian/adaptive search, premium insights & regimes, multi-asset training, educational TV / MT5 / cTrader exports, readiness checks, ~730d history, and 2000 compute tokens / month. Prefer paper; we do not encourage live trading.

Community packs include the full FX major set (EURUSD, GBPUSD, USDJPY, AUDUSD, USDCAD, USDCHF, NZDUSD, EURGBP, EURJPY, GBPJPY), metals (XAU/XAG), crypto (BTC/ETH), Brent, and futures (ES, NQ, CL, GC via Yahoo continuous). Agents retrieve any pack symbol via ensure_community_data / download_market_data, not a three-symbol teaser. Not equities spot, and not a full exchange roll-chain product.

There is no in-app chatbot. You run the AlgoBreeder MCP server and point Cursor, Claude Desktop, VS Code, Codex, or Claude Code at it. Agents call real tools plus get_research_playbook. Premium-gated MCP tools include run_walk_forward, strategy exports, run_multi_asset_training, and premium_session_insights.

Community is free locally (no cloud account): Monte Carlo, sensitivity, anchored walk-forward, and the Dual EMA / MA trend starters. Researcher (€590/yr, ≈ €49/mo billed yearly) is the hosted Lab/MCP edition with Researcher+ validation tools, exports, fuller history, and a monthly token budget — less than TradingView Premium for tools that actually pressure-test edge.

Yes. Community runs locally with research workers, MCP, and the Training Center. After purchase you can also load the commercial Researcher+ package on your own machine; hosted Lab remains the Stripe path for cloud compute.

Cool-down

Answers above. Docs anytime; Researcher plans when you are ready.

After answers

AlgoBreeder

Connect your LLM to a trading research lab for forex, metals, crypto, Brent, and futures. Draft, train, backtest, stress-test — paper first. Not financial advice.

DocsResearcher plans →

Product

  • Workshop
  • Workflow
  • Pricing
  • Agents

Markets

  • Forex
  • Metals · Crypto · Brent
  • Futures

Resources

  • Documentation
  • GitHub
  • Status
  • Security
  • FAQ

Legal

  • Privacy
  • Terms
  • Refunds
  • Risk
  • Cookies
  • Contact

© 2026 AlgoBreeder. Apache-2.0.

Past performance does not guarantee future results. Trading involves substantial risk of loss.