AI Agents
Hub-style LLM blocks. Each accepts connections from messengers, data feeds, trading, files, and more — exposing those as tools the model can call. Connect a Worker or a messenger to give the agent something to do.
Pick the provider you want — the block model is identical, only the underlying LLM changes.
| Block | Type ID | Provider | Description |
|---|---|---|---|
| ChatGPT | ai.llm_agent | OpenAI | GPT models — tasks, prompts, automation. |
| Claude Agent | ai.claude_agent | Anthropic | Claude — vision, reasoning, long-context automation. |
| Grok Agent | ai.grok_agent | xAI | Grok — X/Twitter data, reasoning. |
| DeepSeek | ai.deepseek_agent | DeepSeek | V3 + R1 reasoning, cost-efficient. |
| Groq | ai.groq_agent | Groq | Ultra-fast Llama / Mixtral / Gemma inference. |
Media
| Block | Type ID | Provider | Description |
|---|---|---|---|
| Sora 2 Video | ai.sora_video | OpenAI | Animate a still image into a short video clip (bring-your-own key). |
How agents become useful
An AI Agent on its own just answers prompts. Its power comes from what you wire to it — each connected block becomes a tool the model can call:
| Wire an agent to… | …and it can |
|---|---|
| A messenger (Telegram, Discord, WhatsApp) | Read and reply to messages |
| A Worker | Run arbitrary Python and use any ctx.* adapter |
| A data feed | Pull live market or web data into its answers |
| A trading exchange | Check balances, place orders on instruction |
| A File Explorer | Read and write files in the workspace sandbox |
Turn an agent into a Telegram bot
Wire any AI Agent into a Telegram AI Bot and it answers Telegram messages automatically. See Messengers → Telegram Mini Apps.