Best AI tools for portfolio analysis
PortfolioMetrics has an AI Assistant that runs backtests, optimization and Monte Carlo on any allocation you choose, inside the app, on market data for 500,000+ assets. Account-focused tools analyze the accounts you link, and general assistants write code for each request. The comparison below shows where each fits.
Where PortfolioMetrics is stronger
- The AI Assistant runs backtests, optimization and Monte Carlo on any allocation you choose, inside the app. PortfolioPilot and Mezzi analyze the accounts you link instead.
- Market data for 500,000+ assets comes with the app. General assistants need uploads, code or connectors for a backtest.
- Its backtests matched our reference calculation within 0.5% in the US and 0.1% in Europe.
- AI access starts with a free trial, then $12/month billed annually. testfol.io's AI access needs Max at $80/month billed annually.
Compare at a glance
| PortfolioMetrics | PortfolioPilot | Mezzi | testfol.io | |
|---|---|---|---|---|
| Data | ||||
| Market data | 500,000+ assets, daily | Your linked holdings, daily | Your linked holdings, daily | Daily, through its tools |
| Analysis depth | ||||
| What the AI does | Backtests, optimization and Monte Carlo on any allocation | Portfolio analysis and retirement Monte Carlo | Questions on linked accounts, Monte Carlo on request | Its analysis tools, through an outside assistant |
| Ease of use | ||||
| Where you ask | In the app | In the app | In the app, or Claude and ChatGPT via MCP | ChatGPT, Codex or Claude Code via MCP |
| Model choice | Yes, on paid plans | Not stated | Yes | Your assistant's |
| Price and access | ||||
| AI access from | Free trial, then $12/month billed annually or $14 monthly | $20/month billed annually or $29 monthly (Gold, limited) | Free (limited), then $33.25/month billed annually or $39 monthly | $80/month billed annually or $100 monthly (Max) |
| Trust | ||||
| 60/40 test against our reference | Within 0.5% (US test) and 0.1% (European test) | Not tested | Not tested | Within 0.1% (US and European tests) |
| Scope (not scored) | ||||
| Brokerage accounts | No, enter or upload holdings | Yes, read-only | Yes, read-only | Not listed |
| Personalized advice | No, education only | Yes, SEC-registered adviser, paid plans | Yes, SEC-registered adviser | No |
- Data
- Market data
- 500,000+ assets, daily
- Analysis depth
- What the AI does
- Backtests, optimization and Monte Carlo on any allocation
- Ease of use
- Where you ask
- In the app
- Model choice
- Yes, on paid plans
- Price and access
- AI access from
- Free trial, then $12/month billed annually or $14 monthly
- Trust
- 60/40 test against our reference
- Within 0.5% (US test) and 0.1% (European test)
- Scope (not scored)
- Brokerage accounts
- No, enter or upload holdings
- Personalized advice
- No, education only
- Data
- Market data
- Your linked holdings, daily
- Analysis depth
- What the AI does
- Portfolio analysis and retirement Monte Carlo
- Ease of use
- Where you ask
- In the app
- Model choice
- Not stated
- Price and access
- AI access from
- $20/month billed annually or $29 monthly (Gold, limited)
- Trust
- 60/40 test against our reference
- Not tested
- Scope (not scored)
- Brokerage accounts
- Yes, read-only
- Personalized advice
- Yes, SEC-registered adviser, paid plans
- Data
- Market data
- Your linked holdings, daily
- Analysis depth
- What the AI does
- Questions on linked accounts, Monte Carlo on request
- Ease of use
- Where you ask
- In the app, or Claude and ChatGPT via MCP
- Model choice
- Yes
- Price and access
- AI access from
- Free (limited), then $33.25/month billed annually or $39 monthly
- Trust
- 60/40 test against our reference
- Not tested
- Scope (not scored)
- Brokerage accounts
- Yes, read-only
- Personalized advice
- Yes, SEC-registered adviser
- Data
- Market data
- Daily, through its tools
- Analysis depth
- What the AI does
- Its analysis tools, through an outside assistant
- Ease of use
- Where you ask
- ChatGPT, Codex or Claude Code via MCP
- Model choice
- Your assistant's
- Price and access
- AI access from
- $80/month billed annually or $100 monthly (Max)
- Trust
- 60/40 test against our reference
- Within 0.1% (US and European tests)
- Scope (not scored)
- Brokerage accounts
- Not listed
- Personalized advice
- No
The table shows PortfolioMetrics and the three tools closest to it. Every tool on this page has its own section below. All prices are per month. Billed annually means one payment for the year.
Where other tools are stronger
- PortfolioPilot, Mezzi, ChatGPT, Claude and Perplexity link brokerage accounts read-only.
- PortfolioPilot and Mezzi give personalized advice through SEC-registered advisers.
- Mezzi, testfol.io and everinvest work inside Claude or ChatGPT through MCP. PortfolioMetrics has no MCP server.
- QuantConnect and Composer trade strategies automatically.
PortfolioMetrics
Testing an allocation you choose by asking a question
The AI Assistant runs PortfolioMetrics' backtesting, optimization and Monte Carlo tools on the holdings, dates and settings in your question, then explains the result. AI Insights writes a review of a saved report.
What it does well
- Runs the analysis itself, on market data for 500,000+ assets
- Choice of Claude, GPT and Gemini models on paid plans
- Free trial, no separate AI subscription needed
Check before choosing
- No brokerage linking: you enter or upload holdings
- No personalized advice: results are for education
- No MCP server for outside assistants
Free AI trial. Premium: $12/month billed annually or $14 monthly. Professional: $40/month billed annually or $48 monthly.
PortfolioPilot
Advice on the accounts you already hold
PortfolioPilot links your accounts read-only, backtests the holdings you own today and runs retirement Monte Carlo. Its parent, Global Predictions, is an SEC-registered adviser that gives personalized recommendations on paid plans. The AI Assistant starts on Gold.
What it does well
- Links 12,000+ US and Canadian institutions, read-only
- Personalized advice from an SEC-registered adviser on paid plans
Check before choosing
- AI Assistant is limited on Gold and unlimited from Platinum
- Backtests apply today's holdings to past prices
Free plan. Gold: $20/month billed annually or $29 monthly. Platinum: $49/month billed annually or $99 monthly. Pro: $99/month billed annually or $149 monthly.
Mezzi
AI advice on linked accounts, also inside Claude and ChatGPT
Mezzi links your accounts read-only and answers questions about them in an AI chat, including retirement Monte Carlo on request. It is an SEC-registered adviser. Its MCP server brings the same data into Claude and ChatGPT on every plan.
What it does well
- Personalized advice from an SEC-registered adviser
- MCP server for Claude and ChatGPT on every plan
- Model choice in chat
Check before choosing
- No backtester or optimizer
- Works only with accounts you link or enter
Free plan. Advisory: $33.25/month billed annually or $39 monthly. Concierge: $83.25/month billed annually or $99 monthly.
testfol.io
Deep research with simulated fund history and specialist tools
testfol.io backtests on daily data with simulated tickers that extend fund history back to 1885. It adds factor regression, tactical allocation, leveraged ETF tools and Monte Carlo by historical bootstrap. Its Max plan connects ChatGPT, Codex and Claude Code through MCP.
What it does well
- Simulated history before a fund launched, back to 1885
- Factor regression and historical-bootstrap Monte Carlo
- Full history without an account
Check before choosing
- No AI assistant in the app: AI access needs Max and an outside assistant
- Ticker suffixes for non-US exchanges must be looked up
Free account. Pro: $12.50/month billed annually or $15 monthly. Pro+: $25/month billed annually or $30 monthly. Max: $80/month billed annually or $100 monthly.
everinvest
Free backtests on very long index history from a chat
everinvest backtests index and asset-class portfolios on annual data from 1871, with country tax rules and withdrawal strategies. It has no app of its own: you use it through a Custom GPT in ChatGPT or a free MCP server.
What it does well
- Free, with annual data back to 1871
- Tax rules for 27 countries and 6 withdrawal strategies
Check before choosing
- Annual data, index and asset-class series only, no individual stocks
- No app of its own: needs ChatGPT or an MCP client
Free. No paid plan.
ChatGPT
Open research and custom code on your own data
ChatGPT reads your files, runs Python on them and researches the web with cited reports. It has no built-in price data, backtester, optimizer or Monte Carlo tool, so each analysis is code written for that request. In the US it reads linked accounts read-only through Plaid.
What it does well
- Research and custom code, with a free plan
- Read-only account linking through Plaid in the US
Check before choosing
- No built-in price data, so you upload your own series
- No published methodology for its calculations
Free plan. Go: $8/month. Plus: $20/month. Both billed monthly.
Claude
Documents, research and custom code with financial-data connectors
Claude reads documents, searches the web, runs code and creates PDF, Word and Excel files. It has no built-in price database, so data comes from uploads or from connectors such as Interactive Brokers and Morningstar, each with your own account. Each backtest is code written for that request.
What it does well
- Connectors to providers such as Morningstar, S&P Global and FactSet
- Creates PDF, Word and Excel files from an analysis
Check before choosing
- No built-in price data or backtesting engine
- Account linking reads balances only, in a US beta on Pro and up
Free plan. Pro: $17/month billed annually or $20 monthly. Max: $100/month.
Perplexity
Cited research and market lookups
Perplexity searches the web with citations and has finance data for public equities and ETFs, with price bars from 1 minute to 1 month. Backtests, optimization and Monte Carlo run only as code written for each request in Labs or Computer, on paid plans. It links brokerage accounts read-only through Plaid.
What it does well
- Cited answers, with named finance data providers
- Read-only account linking in the US, Canada and supported European countries
Check before choosing
- Code-based analysis needs a paid plan
- No published list of markets or history length
Free plan. Pro: $17/month billed annually or $20 monthly. Max: $167/month billed annually or $200 monthly.
Julius
Data analysis on your own files
Julius is an AI data analyst. It reads uploaded files and connected databases, runs Python or R and builds reports, presentations and Excel files. Its built-in financial tools pull US equity prices, fundamentals and filings, and each backtest is code written for that request.
What it does well
- Analyzes any data you bring, from files to databases
- US company fundamentals, filings and insider trades built in
Check before choosing
- No built-in backtesting engine
- Built-in market data covers US equities only
Free plan with 25 daily credits. Paid prices are not stated on the pages we could read.
QuantConnect
AI help writing and trading coded strategies
QuantConnect's agent, Mia, writes and backtests strategy code in Python or C#, on data from tick to daily. Strategies trade live through 21 brokerages on paid plans. A remote MCP server opens the platform to outside assistants.
What it does well
- AI agent that writes and backtests strategy code
- Live trading through 21 brokerages on paid plans
Check before choosing
- Strategies are code you need to review
- Not built for allocation research
Free tier. Paid tier prices are not published on the pricing page.
Composer
Building and automating rule-based trading strategies
Composer lets you build rule-based strategies in a no-code editor or with its AI, backtest them on daily data, and trade them automatically through its own broker-dealer.
What it does well
- AI and no-code editor for trading rules
- Automated execution, one strategy on the free plan
Check before choosing
- Trades only through Composer's own brokerage
- Built for trading rules, not allocation research
Free plan. Advanced: $10/month, billed monthly. Pro: $32/month billed annually or $40 monthly.
How we compared the tools
We compared what each tool's AI can do with a portfolio question on five dimensions: data, analysis depth, ease of use, price and access, and trust. Brokerage linking and advice are shown as scope, not scored. Every fact comes from the vendor's own pages, checked in November 2026, or from our hands-on 60/40 test. PortfolioMetrics is listed first because we publish this page.
Frequently asked questions
Yes, when it has market data and a backtesting engine. PortfolioMetrics provides both inside the app. testfol.io and everinvest connect their engines to outside assistants through MCP. General assistants write code for each request on data you supply.
PortfolioPilot and Mezzi link accounts read-only and give personalized advice. ChatGPT, Claude and Perplexity link accounts read-only in some countries. PortfolioMetrics works with holdings you enter or upload.
Not on PortfolioMetrics: results are for education and research. PortfolioPilot and Mezzi are SEC-registered advisers that give personalized advice on their plans. Check each provider's disclosures.
ChatGPT, Claude and Perplexity write code for each request and run it on your uploads, the web or connected sources. Check the data and the method behind each result. PortfolioMetrics runs backtests, optimization and Monte Carlo on daily data for 500,000+ assets inside the app.
PortfolioMetrics lets you choose the model on paid plans. Mezzi, QuantConnect and the general assistants also offer a choice. testfol.io and everinvest use the model of the assistant you connect.
Sources: PortfolioMetrics Features, Pricing, AI tools · PortfolioPilot Pricing, Backtesting, Simulations, Retirement planning · Mezzi Pricing, Features · testfol.io Pricing, Data coverage, AI tools · everinvest Features · ChatGPT Pricing, Account linking · Claude Pricing, Data coverage · Perplexity Pricing, Features, Account linking · Julius Pricing, Data coverage · QuantConnect Pricing, AI tools · Composer Pricing, Methodology. Documentation checked in November 2026. Features and prices can change.
Portfolio analysis is for education and research. Historical and simulated results do not predict future returns.