Building and managing Agents

Last updated: July 13, 2026

Once you know what kind of analysis you want Airspeed to run, the next step is building a saved agent: a workflow with a name, prompt, trigger, output destination, and run history.

This article explains how to create, test, save, edit, duplicate, and manage agents.

Where to start

Open Agents from the left navigation. You can start from several places:

  • Home — run a one-time query from the collaboration space.

  • Discover Agents — choose a prebuilt template and customize it.

  • Command Center — manage your live and on-demand agents, run a one-time query, or create a new agent.

  • Create an agent — open the builder from scratch.

[SCREENSHOT: Agents Command Center showing the Run a one time query and Create an agent buttons]

Start from a template

  1. Go to Agents → Discover Agents.

  2. Browse templates by category or use case.

  3. Select a template.

  4. Run it once, or customize and save it as an agent.

Templates are a good starting point when you want an example prompt, suggested trigger, or recommended workflow.

Start from scratch

  1. Go to Agents → Command Center.

  2. Click Create an agent.

  3. Choose the conversational builder or create manually.

The conversational builder asks what you want the agent to do, then drafts a configuration for you. The manual builder takes you straight to the form.

The builder form

Step 1: Trigger

Choose how the agent starts:

  • Manually — run the agent yourself when needed.

  • Scheduled — run automatically on a daily, weekday, weekly, or monthly cadence.

  • Deal Stage Change — run when a CRM deal changes to a selected stage.

  • After Call — run after a processed call, optionally filtered by call tags.

  • Web Trigger — run when Airspeed detects new web information that matches your monitor query.

[SCREENSHOT: Agent builder Step 1 with the five trigger options visible]

Note: CRM field and deal stage triggers are available for Salesforce and HubSpot. Pipedrive CRM triggers are not currently supported.

Step 2: Prompt

The prompt tells the agent what to analyze and how to format the result. Write the prompt as a clear instruction, including the data scope, time range, criteria, and expected output format.

You can also use:

  • Variables — insert trigger-specific context such as {{deal_id}}, {{call_title}}, or {{monitor_output}}.

  • Web Search — allow the agent to look up public information on the web.

  • Previous runs context — for scheduled agents, include recent previous outputs so the next run can compare changes over time.

[SCREENSHOT: Agent builder Step 2 showing the prompt editor, Variables button, Web Search toggle, and Previous runs context toggle]

Step 3: Output

Choose who receives the result and how it is delivered.

  • Generalized output — the agent runs once, and the same result goes to all selected recipients.

  • Individual output — the agent runs separately for each selected user, scoped to that user's data.

Delivery can include email, Slack direct messages, and Slack channels, depending on your workspace integrations and recipient setup.

[SCREENSHOT: Agent builder Step 3 showing generalized vs individual output options and recipient delivery preferences]

Important: Slack delivery requires the Slack integration to be connected for your workspace.

Test before saving

Use Test Agent to preview the output before saving. If the result is too broad, too narrow, or formatted incorrectly, adjust the prompt and test again.

For triggered agents, the test uses sample trigger context so you can confirm the prompt works with variables and event data.

Save, edit, duplicate, and delete agents

Go to Agents → Command Center to manage saved agents.

  • Live Agents are agents with active automatic triggers.

  • On Demand Agents are saved agents that run manually.

Use an agent's menu to edit, duplicate, delete, or view history. If you are using a shared or global template created by someone else, you may need to duplicate it before you can customize the prompt.

Agent history

Each saved agent has a history of past runs. Open a past run to review the output, continue the conversation, or share the result.

Scheduled and triggered outputs are delivered to the configured recipients, but the in-app history remains the record of what ran and when.

Good first agents to build

  • Weekly pipeline digest — scheduled every Friday, sent to a sales leadership Slack channel.

  • Closed-won handoff — triggered when a deal moves to Closed Won, sent to Customer Success.

  • Post-call follow-up — triggered after discovery or demo calls with selected tags.

  • Competitive news monitor — triggered by web events about competitors or target accounts.

  • At-risk deal alert — scheduled every weekday morning and sent individually to reps.

Related articles

  • :page_facing_up: Airspeed Agents overview

  • :page_facing_up: Using variables in Airspeed Agents

  • :page_facing_up: Using Scheduled Agents

  • :page_facing_up: Agent sharing, history, and collaboration