Set the observation window
Record the start date, simulated starting capital and strategy version before activation. Decide when you will review activity and which condition would make you pause the test. Keep rule changes and cash changes in the log so you can explain the deployed curve.
Paper observation answers a different question
A backtest evaluates past data. A paper portfolio runs the configured strategy forward with simulated money. Use it to inspect the operational behavior, activity and exposure of a frozen candidate; it does not establish real brokerage execution or future profitability.
Create and monitor a separate paper deployment
Deploying an existing live portfolio ID can reactivate real brokerage trading. For this workflow use the newly authored chat candidate and verify the returned deployment is paper. The SDK also separates saved draft IDs from deployment.portfolioId.
| Preview the candidate | build_portfolio | Review the canonical configuration and fix validation issues before saving |
| Author the candidate | create_portfolio | A newly authored chat portfolio, not an existing live account |
| Inspect the target | get_portfolio | Read the exact returned ID and strategies before activation |
| Deploy the chat book | update_portfolio | operations: [{type: "deploy", portfolioId: returnedChatId}] creates paper |
| Find the running book | fetch_portfolios | Keep the resulting deployment ID separately from the original chat/draft ID |
| Read actual paper history | query_portfolio_history | The deployed portfolio curve; never substitute a backtest curve |
No matching rows. Clear the filter to see all records.
Select a newly authored candidate
Connect Claude Code using MCP OAuth. Read the draft’s exact strategy configuration and account identity before activation. Do not recycle an existing live portfolio ID for a task described as paper-only.
Use a newly authored chat candidate for this paper-only test. Read back its portfolio ID and complete strategies. Show the activation plan and wait for approval. Do not deploy, reactivate or replace strategies on an existing brokerage portfolio. After approval, verify the returned deployment is paper and record its new ID.Keep draft and deployment identities separate
SDK save writes a draft; deploy creates the running paper portfolio and returns deployment.portfolioId. Those identifiers coexist. The MCP chat deployment path also creates a separate paper book. Use the running book’s ID for history and monitoring.
Inspect the returned deployment type rather than relying on a portfolio name containing the word paper.
Read the deployed curve
query_portfolio_history reads the deployed book’s actual recorded values. get_portfolio_performance provides aggregates; it is a different output. A historical backtest of the same strategies is not a substitute for the deployed curve.
Keep the observation start date, capital, configuration changes and caveats together. If you stop the paper experiment, target the verified paper deployment identity rather than an unrelated live account.