Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.tedro.io/llms.txt

Use this file to discover all available pages before exploring further.

AI agents are autonomous assistants that handle customer conversations within your workflows. Each agent has a configurable personality, optional knowledge base, and access to tools. You create agents from templates, customize them, test in the playground, and activate them for use in workflows.

​
Creating an Agent

1
Open the Agents Page
2
Navigate to Agents in the sidebar. The agents page shows all your agents as cards with inline stats (total runs, success rate, last active time) and a status indicator.
3
Create from Template
4
Click Create Agent. Select a template that matches your use case — templates provide a pre-configured personality and system prompt that you can customize. Enter a name for your agent and click Create.

​
Configuring Your Agent

The agent detail page has four tabs:

​
Personality

The Personality tab controls how your agent communicates:
  • Name — The agent’s display name
  • Role — Select from predefined roles that set behavioral expectations
  • System prompt — The core instructions that define the agent’s personality, knowledge boundaries, and response style. Write clear, specific instructions for best results.
  • Model — Choose the AI model for the agent (different models offer different tradeoffs between speed, quality, and cost)
  • Active / Inactive toggle — Controls whether the agent is available for workflows
Write system prompts in second person (“You are a customer support agent for…”) with specific instructions about tone, knowledge scope, and when to escalate. More specific prompts produce more consistent results.

​
Knowledge

The Knowledge tab manages the agent’s reference material:
  • Documents — Upload files (PDF, text, markdown) that the agent can reference when answering questions. The agent uses retrieval-augmented generation (RAG) to find relevant information.
  • Links — Add web URLs as knowledge sources.
The agent searches its knowledge base during conversations to provide accurate, grounded responses rather than relying solely on the model’s training data.

​
Actions

The Actions tab configures what tools the agent can use during conversations:
  • Web search — Let the agent search the web for current information
  • File search — Search through uploaded knowledge base documents
  • HTTP tools — Call external APIs (configure URL, method, headers, and body)
  • MCP tools — Connect to Model Context Protocol servers for extended capabilities
Each tool can be individually enabled or disabled. Configure tool credentials and parameters as needed.

​
Performance

The Performance tab shows how the agent is performing:
  • Total runs — How many conversations the agent has handled
  • Success rate — Percentage of runs that completed without errors
  • Response metrics — Timing and quality indicators
Use this tab to monitor agent health and identify issues.

​
Testing in the Playground

Before activating an agent, test it in the Playground:
  1. Open the agent detail page and click the Playground panel
  2. Type messages to simulate a customer conversation
  3. The agent responds using its current personality, knowledge, and tools configuration
  4. Iterate on the system prompt and settings based on the agent’s responses
The Playground uses the same runtime as production conversations, so what you see is what your customers will experience.
Changes to the agent’s configuration take effect immediately in the Playground. You do not need to republish to test updates.

​
Using Agents in Workflows

Once your agent is active, add it to a workflow:
  1. Open the workflow builder and drag an AI Agent node onto the canvas
  2. In the node inspector, select the agent from the dropdown
  3. Connect the node to your workflow — typically after a trigger node and before a send message or handoff node
The agent processes the conversation context and generates a response. The workflow can branch based on the agent’s output or escalate to a human via a handoff node.

​
What’s Next

Quickstart

See how agents fit into the onboarding flow.

Inbox

Monitor agent conversations and handle escalations in the inbox.