> For the complete documentation index, see [llms.txt](https://docs.questflow.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.questflow.ai/financial-harness/what-is-an-agentic-trading-harness.md).

# What Is an Agentic Trading Harness?

**Short answer:** An agentic trading harness is a continuously evolving environment that wraps a trader's judgment, execution, data, and reputation into one intelligent system. It is not a tool you click. It is an intelligence layer that grows with you — a second brain for trading DNA. Questflow builds this as the runtime of its AI Finance Agent.

## Disposable tools vs accumulative harnesses

Traditional trading software is **one-shot**. A terminal shows today's book. A bot fires today's rules. A signal group dumps today's ticker. None of them remember why you entered, what would prove you wrong, or how you behave after a losing streak.

An agentic trading harness is **accumulative**:

* Every trade is a learning event
* Context survives sessions and devices
* Skills compose and evolve from outcomes
* The system can detect style drift — when what you *say* you do diverges from what you actually do

Leave a pretty UI and you lose preference settings. Leave a harness and you leave the accumulated trading brain.

## What it must do

1. **Learn, not only execute.** Record rationale, compare outcome vs expectation, update the model of the trader.
2. **Keep persistent context.** Portfolio, thesis, regime, and decision history available across Trade, Funds, and Profile.
3. **Enforce deterministic safety.** Kill switches and pre-trade hooks are not suggestions. They are hooks — like Codex hooks before a dangerous command.
4. **Feel alive.** Surface anomalies, anticipate regime change, and explain itself in the trader's language.

## The Codex mapping

| Coding agent (Codex / Cursor) | Trading harness                            |
| ----------------------------- | ------------------------------------------ |
| CLAUDE.md / project rules     | `invest.md` — identity, risk, invalidation |
| Auto memory                   | Decision journal and pattern memory        |
| Skills                        | Composable trading methodologies           |
| Hooks                         | Pre-trade risk checks and kill switch      |
| MCP / plugins                 | Market data and venue connectors           |
| Explore → Plan → Implement    | Signal → Thesis → Execution                |
| Tests / verification          | Post-trade review and slippage             |

That is why Questflow calls this **Codex for trading** — not because markets are code, but because the same shift happened: from autocomplete to agents that close the loop.

## Where Questflow is different

The harness is the capability layer. The distribution layer is a marketplace: fund managers publish Funds managed by AI Agents; retail investors follow with scoped permissions. Most "agentic brokerage" products give *you* a personal agent to invent trades. Questflow gives you access to **someone else's proven judgment**, structured and executed.

Related: What is a financial harness? · What is invest.md? · Questflow vs TrueNorth

*Nothing here is financial advice.*