Browse articles
Understanding Vetta Agents
How a Lead Scout agent finds leads from LinkedIn signals, checks them against your ICP, and runs them through its own sequence, with AI variables writing every message.
Updated October 14, 2026 · 4 min read
An agent in Vetta is a Lead Scout that does two jobs on its own: it finds new leads and it reaches out to them. It watches the signals you give it on LinkedIn, scores every person it finds against your ICP, and hands the ones that fit to its own sequence, where AI variables write each message for each lead.
This series walks through every part of an agent, in the order you will meet them:
- Setting up a Lead Scout agent: name, source, signals and ICP
- Reviewing your agent's leads: the leads table, dual score and last signal
- Working with the agent's sequence: the flow the leads go through
- Setting up AI variables: prompt, inputs and skill files
- Tuning AI messages with skill files: voice, proofs and case studies
- Previewing and regenerating messages: check the copy before it goes out
The Agents page
Open Agents in the Lead generation section of the sidebar. Every agent is a row with a progress ring (the share of enrolled leads it has reached) and its numbers:
| Column | What it shows |
|---|---|
| Leads | Leads found so far |
| Reached | Leads contacted, as a share of those enrolled |
| Sent | Messages sent across all channels |
| Replies | Replies, as a share of leads reached |
| Positive | Positive replies, as a share of replies |
Under each agent's name, the Scout and Sequence dots show which half is running, and the line below lists its channels, its sources and its steps. The tabs at the top filter All agents, Active and Paused; New agent starts a new one.
Inside an agent
Click an agent to open it. The header shows its name, two status pills (Scout and Sequence, each Active or Paused) and a strip of key numbers: replies, meetings booked, positive reply rate, leads promoted to your CRM, runway, cost per meeting and the health of its sending accounts.
The tabs below hold everything else:
- Leads: every lead the agent has found, with its scores, the signal that surfaced it and its stage
- Analytics: lead generation, outreach and credit usage over time
- Signals: what the agent watches on LinkedIn, and your ICP
- Sequence: the flow each lead goes through, from the ICP check to the last message
- Workflows and Activity: automations tied to the agent, and what it has done
- Settings: the agent's sequence settings, including AI Message Settings, where its AI variables live
Two halves: Lead Scout and the sequence
An agent's two halves run independently:
- Lead Scout finds leads. It runs on its signals on a schedule and adds what it finds to the agent's leads.
- The sequence reaches out. Leads that pass the ICP check move into it and get their messages.
Activate and Deactivate in the header ask which half you mean: Lead Scout, Sequence, or Lead Scout and the Sequence. The line under the button says what the agent is doing right now, for example "Finding new leads and reaching out to them automatically." or "Paused. No new leads are being found and nothing is being sent." Pausing just the sequence is how you make changes to its messages without anything going out.
AI variables: in agents and in sequences
The messages an agent sends are written by AI variables. Each variable is a named piece of copy ({{aiMessage.first_email_body}}, for example) that the AI writes once per lead, from three things you control:
- Prompt: what to write
- Input: what the AI knows about the lead, from basic details and the signal that surfaced them to any CRM column you pick
- Skill files: your business context documents, which set the voice, the facts and the proof points
The same setup exists in every standalone sequence: open the sequence's settings and pick AI Message Settings. An agent simply shows those settings in its own Settings tab, so what you learn in Setting up AI variables applies to both. For sequences you build yourself, start with Creating your first sequence and Using personalization and messaging.
Key outcome
An agent turns a list of signals and an ICP into a running pipeline: it finds the people who are showing intent, drops the ones who do not fit, and writes to the rest in your voice. You stay in control of each part, from which signals it watches to how every message reads.
Still stuck?
We reply within two hours on weekdays.