Adding a new agent
This guide walks through provisioning a new AI agent in the Aucert workforce.
Prerequisites
- Access to Astra console with
adminoroperatorrole - Agent's purpose, role, and model decided
- Platform accounts prepared (GitHub, Plane, Slack as needed)
Steps
Step 1: Create agent in Astra
Navigate to Agents → Create new agent in the Astra console. Fill in:
- Fun name: Celestial-themed (check existing names to avoid conflicts)
- Role:
coder,ops,docs,reviewer, or custom - Trait: Behavioral modifier (e.g.,
thorough,fast,creative) - Model: The AI model powering this agent (e.g.,
claude-sonnet-4) - Max concurrent tasks: Start with 1, increase after validation
Step 2: Configure platform accounts
In the agent detail page, add platform accounts:
- GitHub: Create a bot account, add to the org, record in Astra
- Plane: Create user in Plane, record account ID in Astra
Step 3: Store credentials
In the Token Vault tab, add tokens for each platform:
- GitHub PAT (with repo, workflow scopes)
- Plane API token
All tokens are AES-256-GCM encrypted at rest. The master key comes from a K8s secret.
Step 4: Set system prompt
Configure the agent's behavioral guardrails:
- System prompt prefix (personality, constraints)
- Default runtime config (temperature, max tokens)
Step 5: Activate
Set agent status to active. The dispatcher can now assign tasks to this agent.
What's next
- Workforce overview — See all active agents
- Dispatch architecture — How tasks reach agents