TheoryCraft vs Composer (2026)
Composer builds an automated strategy you click together and trades it for you. TheoryCraft writes readable Python you validate before any money moves.
Composer is a no-code investing app: you click together a strategy it calls a "symphony", and it auto-trades real money against it on a daily schedule. It is not where you find out whether the idea behind that symphony actually holds up.
You assemble or describe a strategy, Composer turns it into a proprietary symphony (stored as EDN, not code), and rebalances real money against it every day. The backtests behind it run on daily closes, are labelled hypothetical, and benefit from hindsight.
TheoryCraft is built for the research and validation Composer skips. You describe an idea, an AI assistant writes and runs it as readable Python, and the engine replays the market at tick level so you can validate scientifically before you risk a cent.
A Composer symphony only runs inside Composer, and you cannot read or export it as code. TheoryCraft hands you readable, exportable Python you can audit and run anywhere.
TheoryCraft vs Composer: feature comparison
| Feature | TheoryCraft | Composer |
|---|---|---|
| Standard, portable Python (runs anywhere) |
Readable code you can export and run anywhere, with no lock-in
|
Builds a proprietary 'symphony' stored as EDN, not exportable code
|
| Flexible LLM access (included, or your own key) |
Model included on the Managed plan, or your own key across 26 providers
|
Uses its own model only
|
| Open-source engine (Apache-2.0) |
Free to inspect, fork and run
|
Proprietary SaaS (only a thin MCP client is open source)
|
| Self-host the open-source engine |
Run the Apache-2.0 core on your own machine
|
Hosted SaaS tied to its own brokerage
|
| Scientific validation (walk-forward, Monte Carlo, overfit checks) |
Layered Edge, Filters and Risk, each tested independently
|
No walk-forward or Monte Carlo; backtests benefit from hindsight
|
| Tick-level, event-driven backtesting (no look-ahead bias) |
Accuracy by construction
|
Daily-close backtests only, explicitly hypothetical
|
| Historical strategy backtesting |
Across years of market data
|
Backtests strategies on daily bars
|
| Multi-asset markets |
Forex, equities, metals, commodities, indices
|
US assets only
|
| Market data included (1,600+ instruments) |
Forex, EU and US equities, metals, commodities, indices via Dukascopy Bank
|
US equities, ETFs, options and crypto, daily only
|
| Parameter optimization |
Matrix and genetic search
|
No parameter optimization
|
AI on TheoryCraft vs Composer
Composer genuinely has AI, and even an MCP server. You describe an idea in plain English, its AI assembles a symphony (the strategy you would otherwise click together), and an official open-source MCP server lets Claude or ChatGPT create, backtest and run symphonies.
So the question is not whether there is AI. It is what the AI gives you, and what it does with it.
TheoryCraft's AI works toward a different output:
- From a brief to readable Python. Describe the idea and the assistant writes the code, runs the backtest, and reads back the results. You can also write the code yourself; the AI is a head start, not a black box.
- Readable, exportable output. Every strategy is Python in a notebook you can read, audit, export, and run anywhere. Composer's AI builds a proprietary EDN symphony that only runs inside one app and cannot be read as code.
- Research, not auto-trading. TheoryCraft writes code you study and validate. Composer's AI builds bots that automatically trade and rebalance real money.
- Flexible LLM access. Use the model included on the Managed plan, or bring your own key across 26 providers. Composer uses only its own model.
Composer's AI ships a symphony that trades. TheoryCraft's AI writes Python you research with.
Why traders choose TheoryCraft over Composer for research
Readable Python, not a locked symphony
Composer's AI produces a proprietary symphony, stored as EDN, that runs only inside Composer. You cannot read it as code, export it, or run it elsewhere. TheoryCraft generates standard Python notebooks you read, audit, export, and run anywhere. It is the same code a quant would write by hand.
An open engine you can inspect
TheoryCraft's tick-level backtesting engine is open source under Apache-2.0. You can read it, fork it, and run the core yourself. Composer is proprietary SaaS, and only a thin MCP client is open source, never the engine.
Validated at tick level, the scientific way
TheoryCraft replays the market event by event at tick level, so look-ahead bias is avoided by construction, then adds walk-forward and Monte Carlo. Composer's backtests are daily-close only, labelled hypothetical, and benefit from hindsight, with no parameter optimization.
The data Composer lacks
TheoryCraft includes historical tick data for 1,600+ instruments across forex, EU and US equities, metals, commodities, and indices, sourced from Dukascopy Bank. Composer is US-only: US equities, ETFs, options, and crypto, at daily resolution.
Stop guessing. Start proving.
Turn an idea into evidence. Build it, backtest it, and see whether the edge is real before you risk capital.
Frequently asked questions
Is TheoryCraft a good Composer alternative?
Yes, for the research and validation Composer skips. Composer assembles an automated strategy and auto-trades real money on a daily schedule. TheoryCraft turns a plain-language idea into readable, exportable Python, backtests it at tick level, and validates it scientifically before any money is at stake.
Does Composer use AI and an MCP server?
Yes, both. Composer assembles strategies from a plain-language description and ships an official open-source MCP server so Claude or ChatGPT can create, backtest, and run them. The difference is the output: Composer locks the result in a proprietary EDN symphony that auto-trades, while TheoryCraft writes exportable Python you research with.
Can I read and export my strategy as code?
On Composer, your strategy is a proprietary symphony stored as EDN that runs only inside Composer, so you cannot read or export it as runnable code. TheoryCraft generates a real, exportable Python notebook you can read, audit, and run anywhere.
Does TheoryCraft auto-trade like Composer?
No. TheoryCraft is for research and validation and never connects to your broker. Composer, by contrast, automatically executes and rebalances real money daily against a symphony you cannot read. You use TheoryCraft to prove an edge first, then trade it wherever you like.
What market data does TheoryCraft cover that Composer does not?
TheoryCraft includes historical tick data for 1,600+ instruments across forex, EU and US equities, metals, commodities, and indices, sourced from Dukascopy Bank. Composer is US-only and daily: US equities, ETFs, options, and crypto, with no forex or EU coverage.