Agent Configs
Global Prompt Sections
Shared across all configs — per-config can override each section individually
Rules for scheduling, tool usage, and goal detection. Plain text — no Jinja variables needed here.
General tone, language, and behavior rules.
Product info, FAQ answers, and any static knowledge the agent should know.
Criteria the AI uses to qualify or disqualify leads. Tells the agent when to mark a lead as disqualified, not interested, or nurture (interested but not ready now).
No agent configs yet
Create your first configuration to run a campaign.
New Agent Config
Who the agent is — supports
{{ lead.name }}
Just the identity + goal. Scheduling, behavior, and knowledge are inherited from global sections above.
Override global sections (optional — leave blank to inherit)
Conversation Flow
(optional guided script)
Define a step-by-step conversation flow. The AI will follow these nodes based on the conversation history. Leave empty for freeform conversation.
This config uses a legacy full system prompt. Fill in a Persona above to migrate it to the new format, or leave the Persona blank to keep using the legacy prompt as-is.
0 chars
Used when LIFECYCLE_AUTOMATION is off. Manage legacy sequences
Used when LIFECYCLE_AUTOMATION is on for this org. Resolver order: this → org default. Manage automations
Conversation stages
Ordered list of stage keys this agent classifies into (lowercase, comma-separated). Used as a per-agent
Literal for AgentResult.conversation_stage. Leave empty to use defaults.
Using defaults:
initial, qualifying, scheduling, booked
Tools
Which tools the agent can call. Untick to lock the agent out — e.g. negotiator without scheduling tools.
Dynamic context blocks
Per-lead snippets injected into the prompt at runtime. Tick "Quote context" for negotiator-style agents.
Omit static sections
Tick to exclude a section from the agent's prompt. Persona is always included.
Voice agents have their own standalone instructions. The system automatically appends today's date, call format rules, and tool-calling guidance on top of what you write here.
Supports
{{ lead.name }}
Who the agent is, the purpose of the call, qualification questions to ask, and how to handle objections. Write as if briefing a human caller.
Delete config?
Leads assigned to this config will fall back to the default agent.
Voice Test
Browser WebRTC — no SIP call placed
Select a lead above, then click Start Call
Connecting…
AI is ready
You
Transcript
Transcript will appear here…
Tool Events
No tool calls yet
Call ended
Test Agent
Messages saved to lead history — use Clear to reset
Select a lead and send a message
The agent will respond using this config
Enter to send · Shift+Enter for new line