Build Daily
Tinley Park · August 23, 2026Build a social media agent with Managed Deep Agents
This video demonstrates how to build a social media agent using Managed Deep Agents that scans Hacker News and X to generate daily post drafts delivered via Slack. The tutorial covers project setup, custom tool creation, instruction writing, skill development, scheduling, local testing, and production deployment.
The outline
Tap a timestamp to jump there in the video.
Introduction
The creator introduces the social media agent that generates three post drafts every morning using LangChain's Managed Deep Agents.
Project Setup
The video covers installing Python and UV, setting up the MDA CLI, initializing the project directory, and configuring API keys.
Agent Configuration
The creator customizes the agent's model to GPT-5.6 Luna for cost efficiency and defines its core properties in agent.py.
Custom Tools
Tools are created to search Hacker News via Algolia API and fetch X user timelines and posts using LangChain decorators.
Instructions & Skills
The agent's role is defined in instructions.md, while a dynamic skill is created for post drafting with YAML front matter for invocation triggers.
Memory System
A memory.py file is configured to provide cross-thread persistence so the agent remembers previously drafted posts and avoids repetition.
Slack Integration
A Slack channel is set up by creating a channels directory with a Slack.py file to enable direct communication between the user and agent.
Scheduling & Sandbox
A cron schedule is configured for weekday mornings at 9 AM Pacific, and the default sandbox is removed since it's unnecessary for this tool-based agent.
Local Testing
The agent is run locally using MDA dev to test in LangSmith Studio, allowing inspection of model decisions and debugging before deployment.
Deployment & Slack Setup
The agent is deployed to production with MDA deploy, followed by creating a Slack app from the generated manifest and adding environment variables.
Final Testing
The creator tests the deployed agent in Slack by tagging it for post generation and adjusting formatting preferences based on feedback.
A scheduled agent that scans sources and returns three options turns daily content from a blank page into a review task. For solo builders, the value is not automation for its own sake but narrowing the morning decision to choosing among ready-made drafts.