Chatting with your agent

You can chat with your agent from four places, and they all share the same memory and context:

The dashboard at myagentplatform.com
Telegram — fastest on mobile
Discord — good for team environments
Slack — good for work environments

Start a conversation on one, continue it on another. Your agent doesn’t care where the messages come from.

The dashboard chat

This is the richest interface. It supports:

Plain text — type what you want, hit enter
Multi-line messages — Shift+Enter for a new line
File attachments — paperclip icon, drops files into your agent’s workspace
Image paste — paste a screenshot directly from your clipboard
Streaming responses — you’ll see the agent’s response appear as it’s generated, not all at once

The dashboard also shows the tools your agent is using in real time. When it says “Searching the web…” or “Reading your inbox…” you’re watching it actually happen.

On mobile (Telegram, Discord, Slack)

Mobile chat is text-only in most cases, plus images. The agent knows it’s responding on mobile and tends to keep answers shorter — you won’t get 2,000-word responses on your phone unless you explicitly ask for one.

Your agent will also:

Use the platform’s native formatting (Markdown on Telegram, code blocks on Discord, Slack message formatting)
Send attachments as native files (not broken text)
Respect thread context (Slack threads, Discord replies)
Use emoji reactions where appropriate for acknowledgments

What to expect from responses

Your agent is trained to:

Be concise by default. Short tasks get short answers. Long research gets long answers.
Show its work. If it does research, it cites sources. If it runs code, it shows the output.
Ask before acting on anything destructive (sending email, posting publicly, deleting things).
Admit when it doesn’t know. Better than guessing.

It will NOT:

Just agree with you to be nice. It’ll push back if you’re wrong.
Make up sources or data. If it doesn’t have information, it says so.
Act without evidence. If a tool returned an error, it’ll tell you.

Quick actions in the chat header

At the top-right of the chat window you’ll find:

New chat — starts a fresh conversation. Your agent’s long-term memory stays intact; only the current conversation thread resets.
Switch model (BYOK only) — opens the model picker if you’ve added your own API key in Settings → Integrations → Bring Your Own Key. If you haven’t set up BYOK, this is a no-op and your agent keeps running on the platform default.

To approve or reject a pending action, use the Approvals tab in the sidebar — every pending item shows the full context, the requesting agent, and approve / reject buttons.

Conversation length

Your agent doesn’t have a conversation-length limit in the traditional sense — it can keep talking for hours. But it does compress older parts of the conversation when the context gets long, to save tokens. The compression preserves the important stuff (decisions made, facts established, action items) and drops the small talk.

If you want a completely fresh start, click New chat in the chat header. Your agent’s long-term memory stays intact; only the current conversation thread resets.

Copying and regenerating

In the dashboard chat you can:

Copy any response — hover over a message, click the copy icon
Regenerate — if you didn’t like an answer, ask for another attempt: “Try again, more focused on the pricing angle.”

On mobile (Telegram/Discord/Slack), regeneration works the same way — just ask again with the adjustment you want.