---
title: "Coordinating Agent Conversations"
canonical: https://trickest.com/docs/using-the-app/agent/coordinating-conversations
description: "Ask the Agent to consult another of your chats, send a scoped request, and check delivery and results."
---

# Coordinating Agent Conversations

## Before you start

The Agent can find and consult your other ordinary Agent conversations in the same vault. This does not include teammates' chats, archived or deleted chats, workflow-node sessions, scheduled tasks, or helper sessions.

Use distinct chat titles so the Agent can identify the right work. Messages you wrote in those chats can be included as task context, so keep the request scoped to the information needed.

## Ask for context from another chat

Name the conversation and what you want to learn:

> Find my conversation about the staging login check and summarize its latest result, including what remains unverified.

The Agent can read a bounded portion of its text history. A summary may omit earlier context and does not include every tool result. If the answer lacks the evidence you need, open the original chat and inspect the referenced result.

## Send a request to another conversation

Be explicit when you want work to continue there:

> Ask my staging login conversation to verify whether its report is saved, continue that chat if it is idle, and bring back the file link.

The receiving chat keeps its own model, workspace, and permissions. A running chat receives the message at a safe point in its work; sending a message does not create a second simultaneous turn in that chat. A message can also wait for the next turn without waking an idle chat.

## Check delivery and outcome

The transcript shows the sending conversation and the message's delivery state. Expand the message content to inspect what was sent.

<Frame editorial maxWidth="680px" darkSrc="https://res.cloudinary.com/db14crach/image/upload/v1793311600/www/docs/release-1-2-7-2026-10-29/coordination-receipt-dark-62e23ac2ce4e1f7a.png" caption="This expanded test message is marked Delivered. The receipt confirms delivery, not completion of the requested work.">
  <img src="https://res.cloudinary.com/db14crach/image/upload/v1793311605/www/docs/release-1-2-7-2026-10-29/coordination-receipt-light-c0d12d5ad881a43d.png" alt="Expanded message from another conversation with a Delivered receipt." />
</Frame>

| Receipt | Meaning |
| --- | --- |
| Waiting or pending | Accepted, but not yet included in the recipient's context. |
| Added to context or delivered | The recipient received the instruction. This is not proof that the task finished. |
| Cancelled | The message will not be applied; inspect the reason before retrying. |

Confirm the result in the receiving conversation or its saved output. The chat switcher also distinguishes **Running**, **Needs input**, and **Queued**, so you can find work waiting for your attention.

## Handle a blocked conversation

A pending human question or permission request holds message delivery and continuation. Open that conversation and respond yourself. An Agent message cannot supply human approval or override your instructions.

**Stop** cancels pending input for the stopped turn. Failed, archived, and deleted chats cannot be restarted by peer messages. If you want to resume a stopped chat, open it and send a new message yourself. Avoid repeatedly sending the same request while its first receipt is pending.

---
_Markdown source of https://trickest.com/docs/using-the-app/agent/coordinating-conversations._