> For the complete documentation index, see [llms.txt](https://docs.tread.fi/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.tread.fi/agents/starting-and-managing-agents.md).

# Starting and Managing Agents

<figure><img src="/files/QsfxITt9zEPWX71heiey" alt=""><figcaption></figcaption></figure>

Select an agent to run a task.

**To start:**

1. Select agent by clicking the agent card.
2. Send your prompt.

#### Writing a good first prompt

Agents work best when you give them the same details you'd give a human trader. The more of these you include, the less back-and-forth you'll need:

* **Ticker and side** — "buy SOL", "sell BTC perp".
* **Size** — notional ($100,000 worth of BTW) or token count (1BTC).
* **Account** — which whitelisted account to use.
* **Time horizon** — "over the next 2 hours", "by end of day".
* **Constraints** — passive only, don't cross 1%, stop if price moves through X.

Even if the prompt is not complete, the agent will ask clarification questions to gather sufficient information it needs to execute the order.&#x20;

**Reading the chat**

* **Thinking blocks** — the agent's reasoning before it acts. Collapse them if you don't want the detail.
* **Action rows** — discrete actions the agent took or wants to take (e.g. *Place TWAP*, *Cancel order*).&#x20;
* **Chained-order bubbles** — when the agent strings several orders together, they're grouped into one bubble you can expand.
* **Chart bubbles** — inline price charts the agent renders to make a point.

#### Propose Start cards

For anything that puts your money at risk, the agent doesn't just immediately start an order — it builds a **Propose Start** card with the proposed order parameters and waits for your **Start** click. You can edit, reject, or accept.

Some smaller follow-up actions (re-quoting, tightening a spread on a running MM grid) run automatically once the task is live. The mission log records everything so you can audit later.

#### Live status

A small pulse indicator at the top of the chat tells you whether the agent is *Thinking*, *Acting*, *Idle*, or *Paused*. If it goes dark for more than a minute or two while you expected work,&#x20;

### Managing Active Tasks

From any task's chat panel or its row in the left rail:

* **Pause / Resume.** Tells the agent to stop taking new actions without cancelling what's already in flight. Useful when the market just moved and you want a beat to think.
* **Cancel.** Ends the task. If there are no open positions or orders, this is one click.
* **Cancel with positions.** If the task has open exposure, you get the **Cancel With Positions** dialog: choose whether to flatten positions and cancel resting orders, or leave them on the account for you to manage manually. Read this dialog carefully — flattening sends market orders.
* **Edit the brief mid-run.** Just message the agent. "Tighten the spread to 5bps" or "stop after the next fill" — the agent will acknowledge and either act or come back with a Propose Start if the change is material.

#### Review mode

If you see a **Review mode** banner at the top of the chat, the agent has paused itself and is waiting for your input before doing anything else. Common triggers: an unexpected market move, a constraint breach, or a question only you can answer. Reply to clear it.

<br>


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.tread.fi/agents/starting-and-managing-agents.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.