Toggle theme
Agent
Agent β
The Agent tile lets you create a full-fledged intelligent assistant (AI agent) that can analyze requests, follow instructions, use tools, and interact with other elements. You can customize its role, style, tools, and response format to get an intelligent assistant ready to follow your logic.
Beta Feature
The Agent tile, with channel support, multi-layered context, and multi-agent systems, is available in beta and may change in future updates.
What does it do? β
Creates a specialized assistant that acts according to the set instructions and can use connected tools.
When to use? β
When you need to create a chatbot, consultant, documentation assistant, support agent, or an autonomous task handler.
Main tile elements β
Instruction β
Defines the agentβs behavior. Describe what it should do, how to answer, what language to use, and what role to play. Example:
- Answer only based on documentation from example.com, politely decline other questions
- Speak in Engish, style β friendly assistant
- If the question is off-topic β redirect to support@example.com
Function description β
Generated automatically based on the instruction, with the option to edit. In multi-agent scenarios, helps other agents understand when and how to interact with this agent.
Response format β
Choose between:
Markdown
β for formatted textJSON
β ideal for APIs and multi-agent systems where structured responses are important.
Tools β
Allow the agent to perform additional actions:
- Search, image generation, video, speech synthesis, stylization, etc.
You can also attach:
- HTTP functions β for calling external APIs. More: HTTP Function
- MCP services β for connecting to external systems via ready integrations. More: MCP Service
These elements can be used as tools for your agent.
Model selection β
Specify which model the agent will use. More: Choosing a language model.
βApplyβ button β
Saves the agent configuration.
Context, hierarchy, and multi-agent logic β
- The agent sees related tiles with content attached above and can use them as context.
- The agent can access other agents that are above it in the structure, but only one level up.
This lets you build agent hierarchies and break complex logic into modules.
How to use β
- Add the Agent tile.
- Write the instruction: what it does, doesnβt do, and in what style it replies.
- Adjust the description, format, tools, and additional context if needed.
- Choose the model.
- Click Apply.
Channels β
After creating the agent, you can connect it to various communication channels:
Dialog (built-in chat) β
Direct interaction with the agent via the Tila interface. Great for testing and improving agent instructions.
Telegram β
Lets you integrate the agent with a Telegram bot. Create a bot in @BotFather. In the βChannelsβ section, select Telegram: provide the token and a welcome message.
HTTP API β
You can use the agent in any external system via HTTP API. More in HTTP API for agents
Recommendations β
- The agent works within the given instruction. Refine and expand instructions for better results.
- Use Markdown or JSON for precise control over output.
- You can create multi-agent systems, where agents interact β one can call another, pass information, or delegate tasks within the set logic.