Skip to main content

Documentation Index

Fetch the complete documentation index at: https://condor.hummingbot.org/llms.txt

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

Air Date: April 17, 2026 We’re back with Episode 3 of The Bot Pod! This week, Mike and Fede introduce one of Condor’s most powerful features: routines. These lightweight Python files let you automate tasks, fetch external data, and run scheduled reports—all without burning tokens on LLM reasoning.

​
Episode Highlights

​
What Are Routines?

6:00 — Fede breaks down the three key concepts in Condor’s architecture:
  • MCP Tools: Standardized way to present tools for agents
  • Skills: Markdown folders with instructions that load into context when needed
  • Routines: Lightweight Python files that execute deterministically without loading into memory
The key insight: routines are simple, reliable, and don’t waste tokens. They’re just Python code with a config class and an async run() method.

​
Types of Routines

11:00 — Two types of routines exist in Condor:
  • One-shot: Execute once or on a schedule (every 30 seconds, hourly, daily)
  • Continuous: Run forever with a while loop until manually stopped

​
New Condor Features Demo

14:00 — Before diving into routines, Fede shows off recent improvements:
  • Self-update feature: Update Condor directly from the interface
  • New trade page: Deploy any executor type from the web UI
    • Order executor (simple buy/sell)
    • Position executor (with take profit/stop loss)
    • Grid executor
    • DCA executor

​
Hello World Routine

27:00 — The simplest routine demonstration:
  • Configure parameters in the config class
  • Run it once, run it in background, or schedule it
  • Output appears directly in Telegram

​
Top Movers Routine

32:00 — A practical routine that scans Binance for volatile markets:
  • Fetches ticker data directly from exchange API
  • Returns top gainers, losers, and high-volume movers
  • Written entirely by Condor—zero manual coding required

​
Technical Analysis Routine

35:00 — A more sophisticated routine that:
  • Fetches candles via Hummingbot API
  • Calculates trend, volatility, support/resistance
  • Sends charts as images to Telegram
  • Suggests grid parameters based on analysis

​
Live: Building a News Sentiment Routine

40:00 — Mike and Fede build a CoinTelegraph scraper live on stream:
  1. Tell Condor: “Create a routine to parse news from CoinTelegraph”
  2. Condor writes the Beautiful Soup scraping code
  3. Debug and fix dependency issues in real-time
  4. Run the routine to fetch 15 latest crypto news articles

​
News-Driven Trading Demo

47:00 — The complete workflow from news to trade:
  1. Run the news routine to fetch articles
  2. Ask Condor to identify tradeable tokens from the news
  3. Condor identifies NEO based on treasury restructuring news
  4. Run technical analysis routine on NEO
  5. Deploy a long grid based on the analysis

​
Preview: Routines Inside Agents

53:00 — Next week’s episode will cover:
  • Global routines: Available to all agents
  • Agent-level routines: Scoped to specific trading agents
  • How agents can call routines every tick for data processing

​
Resources