Configuring and Training Agents
Set up each agent's model, brain access, campaign assignments, communication channels, and skills - then train them to produce excellent work.
Once you've hired an agent, configuration and training is what turns them from a generic specialist into someone who produces work that matches your standards. Open any agent's profile from the Team page to start.
Model Strategy
Choose how the agent's brain adapts to different tasks:
| Strategy | Best for | How it works |
|---|---|---|
| Economy | Interns and routine tasks | Uses the fastest, most cost-efficient model for simple work |
| Auto (default) | Most team members | Picks the right model for each task automatically. Fast for simple work, powerful for complex work |
| Power | Managers and complex decisions | Uses the most capable model for deep analysis, multi-step reasoning, and strategic thinking |
You can also pin a specific model instead of a strategy. Open the agent's profile, go to the Comms tab, and click the model chip. Strategies are listed first, with specific models below.
Best Practices for Model Selection
| Agent role | Start with | Upgrade to Power if... |
|---|---|---|
| Content writer | Auto | They struggle with nuanced tone or complex briefs |
| Research analyst | Auto | Reports feel shallow or miss connections |
| Developer | Auto | Complex multi-step builds get confused |
| Customer support | Auto | Responses lack empathy or context |
| Finance agent | Auto | Analysis misses edge cases or nuance |
| Social media | Auto | They need to coordinate across platforms |
| Project manager | Power | Start on Power. They coordinate others |
| Brand strategist | Power | Start on Power. Strategic thinking |
Signs you should upgrade: Output feels generic despite good brain content. Agent struggles to connect multiple pieces of information. Complex tasks produce errors.
Signs you should downgrade: Agent is doing simple repetitive work. Speed matters more than depth. The task has very clear, straightforward instructions.
Brain Access
Each agent has toggles that control which brains they can read from:
| Toggle | What it controls | Default |
|---|---|---|
| User Brain | Whether this agent can access your personal knowledge | On (off for support-domain agents) |
| Company Brain | Whether this agent can read org-level institutional knowledge | On (in orgs) |
| Customer Brain | Whether this agent can read real customer signals | On for customer-facing roles |
| Campaign Context | Whether this agent pulls context from assigned campaigns | On |
Support-domain agents have User Brain access disabled by default. This prevents personal business context from leaking into client-facing work like support tickets or customer communications.
When to Toggle Brain Access Off
- Turn off User Brain for agents handling external-facing work where your internal context should not appear
- Turn off Company Brain for client-facing agents that should not leak internal processes
- Turn off Customer Brain for back-office agents that should not see customer-identifiable signals
- Turn off Campaign Context for agents that should work from general knowledge only, not project-specific data
Campaign Assignment
Assign agents to specific campaigns. This controls two things:
- What knowledge they access Agents pull from the Campaign Brain of each assigned campaign
- What work they can do - only assigned agents can execute missions for that campaign
Best Practice: Match Agents to Campaigns Intentionally
Do not assign every agent to every campaign. A content writer working on your SaaS product launch does not need access to your event planning campaign's knowledge. It just adds noise.
Assign agents to the campaigns where their skills and training are relevant. If you need the same role for two very different campaigns, consider hiring two separate agents with different skills and brain training for each. (Jaime can build the second one in one conversation. See Hiring with Jaime.)
Communication Channels
Set up how agents reach you:
| Channel | What it does |
|---|---|
| Telegram | Agent sends updates, asks questions, and shares deliverables via Telegram |
| Slack | Same, through a connected Slack channel |
Configure channels in the agent's profile under Communication.
Daily Digest (CEO and Executive Agents Only)
If your agent is a C-level executive (like Vibey), the Communication tab shows two additional settings:
Default Channel Where the agent sends proactive updates: Team Chat, Telegram, or Slack. This is where Autopilot notifications and status updates are delivered.
Daily Digest A daily summary of all awareness points the agent tracked. Toggle it on and pick a delivery time. The agent sends the digest to your default channel at the time you set.
These settings only appear for C-level agents because they are the ones sending proactive updates and coordinating across your team.
Agent Brain
Give agents their own specialized memory. This is separate from the Campaign Brain. It stays with the agent across all campaigns.
Upload the best content in their specialty:
- A copywriter gets the top copywriters' portfolios and swipe files
- A researcher gets the best analysis frameworks and methodologies
- A developer gets architecture patterns and coding standards
More in Agent Brain.
Skills
Skills are reusable instruction sets that tell agents how to do specific tasks your way. The more specific the instructions and references, the more consistent the output.
Every agent from the library comes with starter skills. Replace them with skills trained on your content and your standards.
Training: The Full Loop
Hiring is step one. Training is what makes agents excellent.
Feed their brain
Upload expert content to their Agent Brain. Quality over quantity. 3-5 world-class references beat 50 mediocre ones.
Test one-on-one
Chat with the agent directly on the Team page. Give them a real task. Review the output.
Give specific feedback
Do not say "make it better." Say "our analysis reports always include a competitor comparison table and end with three actionable recommendations. Here are two examples."
Refine the skill
Update the skill based on what worked and what did not. After 5-10 iterations, the agent is dialed in.
Deploy via missions
Once you trust the output, start delegating through missions. Check deliverables and feed great work back to the Campaign Brain.
Performance and Scoring
Every agent is scored after completing missions across seven dimensions: quality, reliability, initiative, communication, spec adherence, learning rate, and execution speed.
Use scores to identify:
- Which agents need more brain content or better skills
- Which agents are performing well and should handle more complex work
- Whether a model tier upgrade would help
Firing Agents
If an agent is not working out after training, you can fire them from their profile. Their skills and brain content are removed. You can always hire a new agent for the same role and start fresh (Jaime can build a redesigned version in one conversation).
The most impactful thing you can do for agent quality: train their brain and build good skills. A well-trained agent with great skills produces work that matches your standards. An untrained agent produces generic output anyone could have gotten.