Asset Allocation for the Agentic Era.
Build, test, and maintain investment models with your AI agent. ENSEMBLE provides the data, research tools, and infrastructure to keep them running.
$100 credit · 7 days · No card required
01 / Models
Build the strategy you have in mind.
A simple rule or a strategy of strategies. Give your agent the idea. Keep the model, its code and every run in ENSEMBLE.
Explore modelsRank ETFs by 3- and 6-month momentum. Hold the top three above their 200-day trend. Put unfilled positions in short-term Treasuries.
Momentum & trend
- Lookbacks
- 63 / 126 days
- Trend
- 200 days
- Positions
- 3
- VTI34%
- GLD33%
- BIL33%
Inspect source excerpt
# Source excerpt · illustrative logic
score = (prices.pct_change(63) + prices.pct_change(126)) / 2
eligible = score.where(prices > prices.rolling(200).mean())
leaders = eligible.iloc[-1].nlargest(3).dropna()
# Unfilled slots use the chosen defensive asset.02 / Asset intelligence
Find the assets your idea needs.
Ask for the exposure, structure and history your research needs. Your agent gets structured answers, with the gaps still visible.
Explore asset intelligenceFind managed-futures ETFs. Show me their structure, tax reporting and available price history.
- Structure
- RIC
- Tax form
- 1099
- Catalog start date
- 2020-12-02Stored history not verified
- Structure
- RIC
- Tax form
- 1099
- Catalog start date
- 2019-05-08Stored history not verified
03 / Research
Test a variation. Keep the research.
Change a signal, constraint or rebalance rule. Each run keeps its inputs and results, so your next question builds on the last.
Explore model researchRun the model with a 126-day lookback instead of 63. Keep both runs so I can inspect what changed.
- VTI35%
- VXUS25%
- BND30%
- GLD10%
Saved run · 2026-09-23
- Source
- Saved version 1
- Execution
- Next session
- Costs
- Not modeled
- VTI40%
- VXUS20%
- BND30%
- GLD10%
Saved run · 2026-09-23
- Source
- Saved version 1
- Execution
- Next session
- Costs
- Not modeled
Hypothetical backtest records · illustrative weights, not measured returns.
04 / Reusable methods
Use your methods across models.
Have your agent author a weighting method. Save its source, reuse it across models, and see where it’s used.
Explore optimizersWrite an inverse-volatility weighting method. Save it so I can reuse it in my equity and multi-asset models.
Inverse volatility
Weight inversely to realized volatility
Inspect source excerpt
# Weighting excerpt
vol = returns.std()
inverse_vol = 1 / vol
weights = inverse_vol / inverse_vol.sum()05 / Longer histories
Extend history with explicit assumptions.
Choose an older asset or your own model as a proxy. Extend a fund’s research window while keeping synthetic and observed history distinct.
Explore backfillsExtend CTA’s history with WTMF before CTA’s first price. Keep the proxy period distinguishable from observed history.
CTA’s first price2026-09-25
An older managed-futures fund serves as the chosen proxy. Proxy behavior can differ from the fund.
Illustrative history extension · proxy periods are synthetic.
06 / Maintained allocations
Keep allocations current.
Enable scheduled refreshes. Retrieve the latest successful allocation through your agent, or take a broker-format file into your own workflow.
How refreshes workRefresh my model after each supported trading day. Keep the latest successful allocation and the date of its data.
Diversified Core
2026-09-18
- VTI40%
- VXUS20%
- BND30%
- GLD10%
Saved after a successful refresh. Ready for your next session.
Illustrative schedule · refreshes depend on data readiness and your spending cap. Usage is metered; pause anytime.
Your preferred tools
One workspace, wherever you work.
Your agent, scripts and browser share the same models and research.
Room to grow
Start with an idea.
Choose a plan as your research grows. Keep usage within a spending cap you control.
See plans and usage ratesStarter
$0/month
No subscription fee. Pay for usage.
Explore the platformHobby
$33/month
$50 in usage included each month
Develop your strategiesBusiness
$99/month
$200 in usage included each month
Add broker files and moreBuilds and runs use credit, then incur metered charges up to your cap. Reads are free.
Before you begin
A few useful answers.
Do I need to use the web app?
You can research assets, build models and retrieve allocations through a connected AI agent, API or CLI. The web app gives you another way to inspect the same work. Start by creating an account, then connect your preferred tools.
How complex can a model be?
Models support custom Python logic, signals, constraints and adjustable inputs. They can combine other models and reuse your own optimizers. Today’s runtime uses daily data with compute and composition limits; see the methodology for details.
What does it cost?
Start with $100 in credit for 7 days. Hobby is $33/month and Business is $99/month, with included usage. Builds and refreshes are metered, subject to your spending cap. Broker-format files require Business.
Will ENSEMBLE place trades for me?
You review the research and decide what to trade. ENSEMBLE maintains model allocations and can export weights in supported broker formats. Exports do not place orders or confirm broker acceptance. Backtests are hypothetical; past results do not establish future returns.
Give your next idea a place to grow.
Start with $100 in credit. 7 days to explore, no card required.