Bonus: Resources
A curated collection of documentation, articles, blog posts, and videos to deepen your understanding of Deep Agents and the broader agent ecosystem.
Official Documentation
Deep Agents
-
Deep Agents GitHub Repository — source code, examples, and issues
-
Examples Directory — reference implementations including the content-builder-agent
-
CLI Source — the deepagents-cli Textual TUI
Specifications
-
Agent Skills Specification — the SKILL.md format used for progressive skill disclosure
-
AGENTS.md Specification — the memory file convention for persistent agent context
LangChain Ecosystem
-
LangChain Documentation — the agent framework Deep Agents builds on
-
LangGraph Documentation — the runtime that powers
create_deep_agent() -
LangSmith Documentation — LangChain’s observability platform (alternative to LangFuse)
-
LangGraph Cloud — deploying agents as hosted services
Blog Posts
These posts from the LangChain team explain the design decisions and evolution of Deep Agents:
-
Deep Agents — the introductory post explaining what Deep Agents are and the four pillars (system prompt, planning, sub-agents, filesystem)
-
Doubling Down on Deep Agents — v0.2 release introducing pluggable backends, CompositeBackend, conversation summarization, and large tool result eviction
-
Using Skills with Deep Agents — how progressive disclosure works, the SKILL.md format, and integration with the CLI
-
Building Multi-Agent Applications with Deep Agents — the subagent pattern for context isolation and specialization, combined with skills
Podcasts & Videos
-
Talk Python: Deep Agents — LangChain’s SDK for Agents That Plan and Delegate — Sydney Runkle (LangChain) discusses the harness architecture, why docstrings power tool schemas, prompt caching at scale, and non-coding applications of deep agents. The episode takeaways are an excellent standalone reference covering all key concepts.
Context Engineering & Agent Architecture
Understanding how context fails is essential for building reliable agents. These articles explain why subagents, summarization, and context isolation matter:
-
How Contexts Fail and How to Fix Them — names four failure modes (poisoning, distraction, confusion, clash) with concrete examples from production agent systems
-
Context Engineering for AI Agents: Lessons from Building Manus — practical lessons on KV-cache management, tool masking, pattern mimicry, and treating the filesystem as external memory
-
Context Engineering — Simon Willison’s overview connecting the emerging discipline to practical agent development
Observability & Tracing
-
LangFuse Documentation — open-source LLM observability platform
-
LangFuse LangChain Integration — the CallbackHandler used in the bonus tracing module
-
LangFuse Self-Hosting Guide — run LangFuse locally with Docker
-
LangSmith — LangChain’s native tracing and evaluation platform
Model Providers
Deep Agents supports any LiteLLM-compatible provider. Here are the ones covered in this workshop:
-
Anthropic API Documentation — Claude models (the workshop default)
-
OpenAI API Documentation — GPT models
-
Ollama — run open-source models locally
-
LiteLLM Provider List — full list of supported providers and model formats
Related Projects & Tools
-
uv — the Python package manager used throughout this workshop
-
LangGraph Studio — visual debugger for LangGraph agents
-
Showroom Template — the Antora-based lab format this workshop is built on
Community
-
Deep Agents Issues — report bugs and request features
-
Deep Agents Discussions — community Q&A and ideas
-
LangChain Discord — real-time community chat
Key Concepts Reference
A quick lookup for the core concepts covered in this workshop:
| Concept | Where to learn more |
|---|---|
|
Module 1 and source code |
Four Pillars |
|
SubAgent TypedDict |
|
SKILL.md format |
|
AGENTS.md |
|
|
|
Middleware stack |
Module 8 and source code |
Backends |
|
Human-in-the-loop |
|
LangFuse tracing |