Agents
What you configure on an agent after it is created
An agent's model, access and abilities change at any time and independently of each other. All of it lives on the agent's page, laid out across tabs.
The tabs of an agent's page
- General: name, description, instructions, type, key, the "Enabled" switch.
- Chat: a conversation with the agent right in the dashboard.
- Models: which model answers for which purpose.
- Channels: what the agent is talked to through from outside, Telegram included.
- Skills: what the agent can do and how it learns it.
- Connections: what it can reach.
- Files, Tool Calls, Runs: what is left behind after the work.
An "External AI" agent has fewer tabs: no chat, no channels, no runs and no models; it comes for the tools itself.
The order
- Your own LLM provider: to leave the free tier's limits behind and pick the model yourself.
- An agent's connections: what the agent is allowed to touch outside.
- An agent's skills: how it should do the work.
- Editing an agent: name, prompt, key, disabling, deletion.
Agent types only matter if the agent's brain lives on your side.