Skip to content
TradeTwinYour strategy. Your AI twin.
Learn
ไทยEN
Sign in↗
TradeTwin/Learn/How do I bring my own strategy into an AI trading bot?

Your strategy

How do I bring my own strategy into an AI trading bot?

A practical path from your trading explanation to a reviewed proposal: guided chat, audio knowledge, manual MCP and exact plan consent.

TradeTwin · Product educationPublished October 4, 2026

THE SHORT ANSWER

Start by explaining your method, its exceptions and its risk boundaries. TradeTwin's private onboarding can help draft a proposal, while audio knowledge supplies context and a compatible external AI can propose through MCP. These are separate inputs. You review the exact plan before accepting it; none of them automatically starts trading.

Describe decisions, not just a setup name

A strategy label alone leaves important choices unresolved. Explain which observations matter, what confirms an entry, what invalidates it, how exits are decided, and when to stay out. State the market and account you intend to use. Distinguish instructions you actually follow from ideas you are considering. The assistant should expose missing information so the proposal can be reviewed, rather than treat every confident sentence as an executable command.

Use the interview at your own pace

In a configured private workspace, choose Thai or English, or specify another supported language code for the interview. The assistant can ask about your method and boundaries. You may answer that you do not know. Enough requests a draft from the available information, and stopping an answer does not authorize an order. Review assumptions carefully: an unanswered sizing question must remain unresolved instead of becoming an unapproved risk limit.

Keep voice knowledge and strategy proposals distinct

Customer knowledge intake is audio only. Record an explanation and inspect the transcription once the private processing and retrieval setup is ready. A transcript gives the assistant context; it is not a consented trading plan. A strategy document follows a different route: you can export text and manually ask your external AI to read your chosen material and create a proposal. Direct document, text and URL imports into the customer KB are not offered.

Connect your own AI through MCP

MCP provides a structured way for a compatible external client to call the private proposal tools. The current private connection uses a manually supplied Bearer token, with expiry and revocation. The tools can read a strategy contract, retrieve your proposal and submit a draft. They cannot accept the plan, access broker secrets or place trades. The portal's connection card requires a verified private workspace; a workspace request alone supplies no working endpoint.

Review changes before accepting them

Read the full strategy, assumptions and risk fields in the private UI. Acceptance is tied to the exact proposal version and hash, so an edited or replacement proposal cannot silently inherit the earlier decision. Confirm that the written method can be represented by supported execution rules. Account verification and native protection remain separate. Manually importing your selected knowledge does not create an automatic copy feed or make another person's historical results your own evidence.

Questions worth asking

Can my external AI approve a plan through MCP?+

No. The existing private MCP tools are proposal tools, with no consent or order command. Review and acceptance happen in the private owner interface. Trading activation requires additional account, rule and protection checks.

Can I upload a PDF directly as customer knowledge?+

Customer KB intake accepts audio, not direct PDF, document or URL imports. You may manually ask your own external AI to read chosen strategy text and propose it through a compatible MCP connection, then review the result.

Does every ChatGPT or Claude client connect automatically?+

No universal compatibility is claimed. The client must support the private endpoint's HTTP protocol and manual Bearer headers. OAuth and universal portal-to-workspace dispatch are not available in this pilot; private login and provisioning are separate.

Primary documentation for further reading

  • Model Context Protocol: Streamable HTTP↗

API documentation describes the provider's interface. It does not certify your account or TradeTwin connector as ready to trade.

CONTINUE LEARNING

The conceptWhat is an AI trading second brain?↗Accounts and costsService credits, BYOK and trading funds: what is the difference?↗Customer controlWhy do risk limits and exact plan consent matter?↗
All education articles© 2026 TradeTwin · Public pilot · registration does not start trading