Deep Agents in Action
A hands-on guide to building production-grade AI agents with the LangChain/LangGraph ecosystem.
- Source repo
- datawhalechina/deepagents-in-action
- Stars
- ★ 2.7k
- Last updated
- today
- Primary language
- Jupyter Notebook
- FA score
- 39/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformOpenAI API · Claude API
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- Developers wanting a systematic introduction to building AI agents with LangChain/LangGraph, with hands-on tutorials.
- Not a fit if
- Developers stuck on Node.js below 22.12.0 who cannot upgrade their environment
- Readers who want docs only and will not install AgentSeek or configure .env
- Source review
- 39/100 · Major gaps 8 safety controls not found
What does this agent do, and when should you use it?
Deep Agents in Action is a Chinese-language course by Canghai Jiuzhou (LangChain Official Ambassador), teaching how to build production-grade AI agents using the LangChain/LangGraph ecosystem. The course covers virtual filesystems, task planning, sub-agent orchestration, Skills, long-term memory, human-in-the-loop, sandboxing, filesystem permissions, and MCP. The course website is built with Astro and hosted on GitHub Pages, with companion video and article collections on Bilibili and Xiaohongshu. Code examples default to using models via the SiliconFlow cloud platform, with flexible model selection via environment variables. The course text is licensed under CC BY-NC-SA 4.0, and the website source code is MIT-licensed.
The course provides a series of Markdown chapters (in content/) that are preprocessed by scripts/prep-content.mjs to inject frontmatter, then built into a static site by Astro. It guides learners through using the AgentSeek toolkit to create a DeepAgents template, install npx skills for development, and run code examples that demonstrate deepagents features like virtual filesystems, task planning, sub-agents (including async sub-agents), Skills, LongTermMemory, Human-in-the-Loop, sandbox execution, and MCP integration. Examples use the MODEL_NAME environment variable to manage models, defaulting to Qwen or DeepSeek models via SiliconFlow.
- Developers wanting a systematic introduction to building AI agents with LangChain/LangGraph, with hands-on tutorials.
- Teams needing to implement task planning and decomposition for complex tasks, based on Chapter 4.
- Developers needing to execute code safely inside an agent, using the sandbox content in Chapter 10.
- Developers wanting to restrict an agent's filesystem access, following Chapter 11 on filesystem permissions.
- Teams looking to extend agent tool ecosystems via the MCP integration covered in Chapter 12.
How do you install or deploy this agent?
Local development requires Node.js >= 22.12.0. Run npm install in the repository root to install dependencies.
How do you use this agent?
- Run npm run dev to start the development server (with content preprocessing). 2. Run npm run build for production build, and npm run preview to preview. 3. Edit Markdown chapter files under content/ and update metadata in scripts/chapters.json. 4. To run code examples, sign up at SiliconFlow, configure an API key, and set the model via the MODEL_NAME environment variable.
What are this agent's strengths and limitations?
- Authored by a LangChain Official Ambassador, ensuring credible and up-to-date content.
- Well-structured curriculum covering cognitive, core, and advanced topics comprehensively.
- Multiple learning resources (Bilibili videos, Xiaohongshu articles, course website) cater to different preferences.
- Content targets Deep Agents >= 0.5, with some advanced features requiring higher versions.
- Examples default to the SiliconFlow platform, which may inconvenient users preferring other providers.
- Local development requires Node.js >= 22.12.0, excluding those on older versions.
How does this agent compare with similar options?
This course focuses on Deep Agents (based on LangChain/LangGraph), differing from other LangChain tutorials by diving deeper into production-grade agent details.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Deep Agents in Action This agent | 39 · Major gaps | CLIFree + model costs | ★ 2.7k | today | Jupyter Notebook | OpenAI API · Claude API |
| Dive into LangGraph | 69 · Some gaps | Library / SDKFree + model costs | ★ 461 | 1mo ago | Jupyter Notebook | Claude Code |
| AI Agent Engineering Textbook | 55 · Major gaps | FrameworkFree | ★ 601 | 2d ago | HTML | — |
| AI Agents — The Definitive Guide (Companion Code) | 45 · Major gaps | Library / SDKFree + model costs | ★ 2.8k | 2mo ago | Jupyter Notebook | — |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows this is a course website repository without executable Agent code, so all trust-related criteria are not applicable and scored 0. No evidence of permission management, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, rollback, or source attribution.
Self-consistency: README and package.json are consistent, project structure is clear, score 2. Dependency availability: dependency list is explicit, but no lock file or integrity checks, score 2. Failure messages: asset validation scripts exist, but no user-facing error messages, score 1.
Audience and scenarios: clearly targets developers learning Deep Agents, provides chapters and scenarios, score 2. Capability boundaries: no explicit scope or limitations, score 1. Trigger precision: no trigger mechanisms involved, score 1. Environment fit: provides Node version requirement and installation steps, score 2.
Information architecture: README provides clear table of contents and structure, score 2. Install notes: provides npm install and start commands, score 2. Naming stability: version 0.0.1, but no naming conventions, score 1. Examples and FAQ: provides chapter examples, but no FAQ, score 2. Known limitations: not explicitly listed, score 1. License: clearly CC BY-NC-SA and MIT, score 2. Versioning and changelog: no changelog, score 1. Maintenance responsibility: contribution guide and PR welcome, score 2.
Output usability: course content is well-structured and directly usable, score 2. Marginal value: provides unique Deep Agents practical content, score 2. Cost-benefit: free and open source, but depends on external model services, score 2.
Claim traceability: some claims have links, but no verification methods, score 1. Cross-source corroboration: external links exist, but no cross-validation, score 1. Fact-inference separation: not clearly distinguished, score 1.
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Not found in source: verifiable attributionInstall from the official repo or registry and check the publisher and URL to avoid look-alike packages.
- This repository is a course website without executable Agent code, so trust-related criteria are not applicable.
- Dependencies are not locked, posing supply chain risks.
- No known limitations or changelog provided; users need to assess themselves.