Core Concepts
Agents
An Agent is the core entity in Talknex. It defines how your AI behaves, what it sounds like, and what it can do during a call.
Agent anatomy
Every agent has five configuration areas:
Writing effective system prompts
The system prompt is the most impactful configuration. A well-written prompt produces consistent, on-brand responses. A poorly-written one leads to hallucinations and off-script behavior.
Best practices:
- 1State the agent's role and company in the first sentence.
- 2List what the agent CAN and CANNOT do explicitly.
- 3Specify how to handle edge cases: angry callers, out-of-scope questions, silence.
- 4Include escalation instructions: "If the caller asks for a manager, say X and transfer."
- 5Keep it under 800 tokens. Longer prompts add latency.
Built-in tools
Every agent has two built-in function tools that the LLM can call during a conversation:
Checks calendar availability and creates an appointment. Requires Google Calendar or Outlook to be connected. Returns a confirmation number that the agent reads back to the caller.
Warm-transfers the live call to a destination phone number. The AI session ends immediately after transfer. Configure the destination number in the agent's Transfer settings.
You can add custom function tools under Agent → Developer → Custom Tools. See the LLM Configuration docs for schema format.
Agent limits by plan
| Plan | Agents | Concurrent calls | Custom tools |
|---|---|---|---|
| Free | 1 | 1 | — |
| Pro | 25 | 10 | Unlimited |
| Enterprise | Unlimited | Unlimited | Unlimited |