Create Your First Agent

This guide walks you through creating a complete agent from scratch — from naming it to deploying it in your first meeting.

→ Open the Agents dashboard →

Step 1 — Open the Agent Builder#

From the dashboard, click Agents in the left navigation, then click New Agent (or the + button).

The Agents list page showing the "New Agent" button in the top-right corner, with an empty state or a list of existing agents below

Step 2 — Name your agent and choose a colour#

Give your agent a name that describes its purpose — for example, Sales Assist, Objection Coach, or Compliance Monitor. The name is displayed in the dashboard, shown to other meeting participants when the bot joins, and used in post-meeting summaries.

Choose a colour to help distinguish this agent visually in your dashboard, especially if you run multiple agents.

The top of the Agent Form showing the name field and colour picker

Step 3 — Set a Persona Prompt (optional but recommended)#

The persona prompt tells the AI who it is and how it should behave. Think of it as the agent's system prompt — a paragraph or two that sets the tone, context, and constraints.

Example persona prompt for a sales agent:

A strong persona prompt will produce more accurate, on-brand responses and fewer irrelevant triggers.

The Persona Prompt text area in the Agent Form, showing an example prompt filled in

Step 4 — Choose Live, Hand raise, or Silent mode#

This is the most important decision for your agent.

Live mode — the agent speaks aloud in the meeting using an AI voice. Other participants can hear it.

Hand raise — the agent creates it's audio response, then signals through the video feed that it has a response ready. Multiple responses can queue for each agent. When ready, call the agent by name and the agent speaks aloud in the meeting using an AI voice. Other participants can hear it.

Silent mode — the agent sends insights privately to a back-channel: your dashboard, a Slack message, a Microsoft Teams message, or a Telegram chat. No one in the meeting hears anything.

You can always change the mode later. If you're not sure which to pick, start with Silent and route to your dashboard — it's the lowest-friction way to see your agent in action.

→ Learn more: Live vs Silent Agents →

The mode selector in the Agent Form showing "Live" and "Silent" toggle options with brief descriptions of each

Step 5 — Choose a Voice (Live agents only)#

If your agent is in Live mode, choose a voice from the ElevenLabs voice library. Voices are previewed directly in the form so you can hear how they sound before committing. There are additional voices available by clicking to view more.

The Voice Selector showing a grid of available voices, each with a play button to preview

→ Learn more: Voice and Persona →

Step 6 — Set Sensitivity#

The Sensitivity slider controls how readily the agent fires.

• Low — the agent waits for very clear, unambiguous matches before triggering. Fewer false positives, may miss subtler cues.
• Medium (default) — balanced. Recommended starting point for most agents.
• High — the agent triggers on loosely related mentions. More coverage, higher chance of irrelevant responses.

Start at Medium and adjust based on your experience.

The Sensitivity slider showing Low / Medium / High options

Step 7 — Configure Back-Channel Delivery (Silent agents)#

If your agent is in Silent mode, choose where insights are delivered:

• Dashboard — insights appear in your CoAgentor meeting event log (default)
• Slack — delivered as a DM or channel message to a connected Slack workspace
• Microsoft Teams — delivered as a DM or channel message to a connected Teams tenant

You'll need to connect Slack or Teams first if you haven't already.

→ Connect Slack → → Connect Microsoft Teams →

The Back-Channel section showing the Slack DM/Channel pickers after a Slack workspace has been connected

Now I have the full article format, voice, and structural conventions. Writing the section to slot between Step 7 (Back-Channel Delivery) and Step 8 (Trigger Rules), becoming the new Step 8 with triggers and post-meeting summary shifting up to Steps 9 and 10. mdx---

Step 8 — Add Data Contexts (optional)#

Data contexts give your agent access to your uploaded files during meetings. When a trigger fires, CoAgentor searches the context's files for content relevant to the live conversation and injects the most useful passages into the agent's response — so it can answer questions or reference specific material without you needing to brief it in real time.

You can add contexts now during setup, or come back and add them later from the agent's edit page.

Adding a context#

Click Add context to open the context builder.

1. Name — give the context a descriptive name, e.g. Sales Playbook, Product FAQs, Competitor Matrix
2. Description (optional) — a short note about what the context contains
3. Color — used for visual identification in your contexts list
4. Files — select files from two sources:

• Uploaded files — any files you've previously uploaded to Data Sources. Files marked as Processing can be selected now. Files marked Failed cannot be selected.
• Google Drive — if you've connected a Google Drive account, expand the Drive section to browse your folders and select files directly. Drive files are downloaded and indexed automatically when you save the context.

5. Scope — controls which meetings the context is active for (see below)

The Add Context dialog in the agent form, showing the name field, file picker, and scope selector

Scope options#

| Scope | When it's active | |---|---| | This agent only | All meetings where this agent is present | | All agents (account-wide) | Every meeting across your account, regardless of which agent is deployed | | This agent in a specific meeting | Only for this agent in one upcoming meeting you select |

This agent only is the right choice for most contexts — it keeps the knowledge focused on the agent that needs it. Use account-wide for general baseline knowledge every agent should have access to (e.g. company product documentation). Use the specific meeting option when you want to load one-off meeting prep material — a deal brief or a proposal — that shouldn't carry over to other calls.

You can add multiple contexts to a single agent. Their contents are retrieved additively — all matching contexts are searched when a trigger fires.

If you're in create mode#

If you haven't saved the agent yet, contexts you add here are held as pending drafts and displayed with a Pending badge. They'll be created and attached automatically when you click Create Agent at the end of this form.

The Data Contexts section in the agent form showing two pending context drafts before the agent is saved

→ Deep dive: Contexts and Scopes → → Upload files to your data sources →

Step 9 — Add Trigger Rules#

Trigger rules are the heart of your agent. Each rule tells the agent:

1. What to listen for — a plain-English description of the condition
2. What to do — an instruction for how the agent should respond
3. How to deliver it — speak aloud (live agents) or send silently (silent agents)

Click Add Trigger Rule to open the rule builder.

Example rule: Competitor mentioned

• Label: Competitor mentioned
• Description: When a competitor product or company is mentioned by name
• Response instruction: Briefly summarise our two key advantages over this competitor in two sentences. Be factual and avoid disparaging language.
• Action type: Speak (if live) or Silent (if silent)

You can add as many rules as your plan allows.

The Trigger Rule Builder panel showing a fully filled-in rule with label, description, response instruction, and action type selector

→ Deep dive: Trigger Rules →

Step 10 — Enable Post-Meeting Summaries (optional)#

Toggle Post-Meeting Summary on to receive a summary after every meeting where this agent was active. The summary includes the full event log — every trigger that fired, the transcript context that triggered it, and the agent's response.

Summaries can be delivered by email, to the dashboard, to Slack, or to Teams.

The Post-Meeting Summary toggle and delivery channel selector in the bottom of the Agent Form

Step 11 — Create Agent#

Click Create Agent. Your agent is now ready to deploy.

Click "Create Agent" to save and create the agent

→ Now deploy it in a meeting →

What happens next?#

• Your agent appears in the Agents list with its mode, colour, and last-used date
• You can assign the agent to a calendar event (if calendar auto-join is configured) or start a meeting on-demand
• Every meeting creates a snapshot of your agent's config at that moment — so you can always see exactly what rules were active for any past meeting

The Agent version history showing all the edits and changes to an agent, with the ability to revert to a previous version.