AI agents in Pod are the shortcut you use when you want help with a specific deal right now — draft a follow-up email, summarize what changed since your last meeting, flag risk signals, pull together a deck for tomorrow's QBR, push an update into your CRM. Instead of opening a new tool for each of those tasks, you pick an agent on the deal you're already looking at, and it goes.

This article covers how to open the agent picker, choose or import the right agent, launch it on the current deal, and keep working in the same thread after the first answer comes back. The workflow is the same in Account 360° in the Pod web app and in the AI Agent tab of the Chrome Extension.

If you want to build a new agent from scratch, start with Build a new agent. If you want an agent to run automatically on a schedule or a trigger, see Set up agent triggers and Use the Agent Feed. If you want to give an agent new tools to act against external systems, see Connect MCP servers and Manage agent actions.

How to access the agent surface

In Account 360° (Pod web app)

Open Pod.
Open a deal in Account 360°.
Stay in the Deal Coach view.
Find the Deal Agent section.

In the Chrome Extension

Open the Pod Chrome Extension while viewing a deal in Salesforce or HubSpot.
Open the deal record in the extension.
Select the AI Agent tab.

The AI Agent tab in the Chrome Extension is available on deal records only.

Three ways to open the agent picker

Once you're on the Deal Agent / AI Agent surface, there's one picker that lists all the agents you can launch — the agent browser panel — and three ways to open it:

Type / in the input. Fastest when your hands are already on the keyboard. A leading / opens the panel; anything you type after the / filters the list.
Click the + button in the input's action row. Fastest when you want to browse without typing — Pod inserts / for you and opens the panel.
Click Explore Agents. Available as a prominent button in the center of the empty agent view and as an icon in the expanded and collapsed headers. Opens the same panel.

All three paths land on the same place, so pick whichever matches what you're already doing.

What's in the agent picker

The panel shows your own agents, grouped top-to-bottom in this order:

Recently Used — the last few agents you launched, newest first. This group appears once you've launched at least one agent; it's how you re-run yesterday's agent in one click.
Personal Agents — agents you built for yourself.
Org Agents — agents built for your workspace by an admin; everyone in the workspace can see and run them.

Each row shows a colored agent icon, the agent's title, its full description, and a small badge on the right telling you whether it's a Personal or Org agent.

Pod's pre-built agents are not listed in this picker. They live in the Library and have to be imported into your workspace before they can be launched — see the Library section below.

Launch a saved agent

From the picker:

Click a row, or use Arrow Up / Arrow Down and press Enter to select one.
The / and any filter text you typed are cleared from the input.
The agent runs on the deal you're currently viewing.
The picker disappears and the conversation thread takes its place.

You don't have to type a prompt of your own — the agent knows what to do based on how it was built.

Filter the list to find an agent fast

Once the picker is open, keep typing after the / and the list filters in real time:

Matches on both the agent's title and its description (case-insensitive).
Groups collapse to show only matches. If nothing matches, you'll see No agents match your search.
Press Escape to close the picker without launching anything, or press Backspace until you clear the / and the picker closes on its own.

Just ask a question instead

You don't have to use a saved agent at all. If you type directly into the input without a leading /, Pod answers your question against the current deal's context using the default Deal Agent.

This is the right path when:

Your question is a one-off and doesn't justify saving an agent.
You want to probe the data before you decide what kind of agent to build.
You want a quick follow-up in plain language in the middle of an existing conversation.

Pre-built agents and the Library

Pod ships a catalog of pre-built agents — templates for the workflows we see most often across customers. You browse and import them from Settings → Agents.

Open Pod in the web app.
Go to Settings → Agents.
Switch to the Library tab.
Browse by category (the chips across the top) or search by name or description.
Click any card to preview the agent. The drawer opens in read-only mode so you can see the title, description, prompt, response format, and whether web search is enabled.
Click Import on the drawer. Pod reopens the agent builder prefilled with the template's content so you can adjust it if you want — rename it, tweak the prompt, change the response format, flip web search.
Save to add it to your workspace.

After import, the imported agent lands in your Your Agents tab as a new Personal agent (admins can import it as an Org agent for the whole workspace). From that point on, it behaves exactly like any other agent you built — it shows up in the agent picker on every deal, and you launch it the same way.

You cannot launch a pre-built agent directly from the picker. Importing is the one-time step that brings the template into your workspace; after that, the picker is the only place you need.

Continue the conversation

After the first answer appears, keep working in the same thread:

Type a follow-up in the input to keep the conversation going. The agent uses the previous turns and the deal's current context for its next answer.
Click New agent to clear the current conversation and start a fresh one — useful when you want to run a different agent without the previous thread bleeding into it.

Open history on a deal (Account 360°)

If you want to revisit an older agent run on the same deal:

Open the Deal Agent section in Account 360°.
Click History.
In Deal Agent History, select the earlier run you want to reopen.

The Chrome Extension uses the Pod web app as the canonical history surface — open Pod in a new tab and follow the same steps there.

Edit, retry, or react to the latest answer

Once a response is complete, you can manage the latest run directly in the thread:

Click Edit on your latest prompt to revise and resubmit it.
Click Retry to run the latest prompt again.
Use Thumbs up or Thumbs down to leave feedback on the latest answer. Feedback helps the team tune default prompts and catch drift in agent quality.

In the Chrome Extension

The agent picker works the same way in the Extension as in the web app for the / trigger, the + button, and the Explore Agents button in the collapsed header. One small difference to be aware of: when the Deal Agent section is already expanded, the Explore Agents icon in the header opens a standalone agent modal rather than the slide-up picker. The modal shows the same agents; it's just a different container.

A few more Extension-specific notes:

The AI Agent tab is available on deal records only.
Settings → Agents (including the Library) and Settings → MCP Servers are not reachable from inside the Extension. When you hit a link that goes there — for example, the Create Agent button in the empty state — Pod opens in a new browser tab at the right page.

Empty states and troubleshooting

No agents yet. If your workspace hasn't built any Personal or Org agents, the picker shows No agents yet — Build your first custom agent to get started with a Create Agent button. In the web app the button takes you to Settings → Agents; in the Chrome Extension it opens Pod in a new tab at the same place.
No results for your search. If your filter text doesn't match any agent titles or descriptions, the picker shows No results — No agents match your search. Clear the filter or press Escape.
Failed to load agents. If the agent list couldn't load, the picker shows a short error with a Retry link. Click Retry; if it keeps failing, send us a message.

When to use this page vs. other AI agent pages

Use this page when you want to run or continue agent work on a deal right now.

Go to these related pages when your goal is different:

Use Build a new agent when you want to create or edit a saved agent.
Use Manage agents when you want to organize saved agents or their automated triggers from Settings → Agents.
Use Set up agent triggers when you want an agent to run automatically.
Use Use the Agent Feed when you want to review results from automated runs in one place.
Use Manage agent actions when you want to approve, reject, or troubleshoot the specific actions agents take on your behalf.
Use Connect MCP servers when you want to give your agents access to additional external tools.
Use the AI Agent Best Practices Guide when you want help writing stronger prompts or response formats.