Skip to slide
127 / 170

CH 14 · Autonomous Task Agents: Manus and CodeAct · 01/11

Read as one page

Autonomous Task Agents: Manus and CodeAct

CodeAct has the model write a short Python program as its action instead of a fixed function call. Manus keeps a todo.md file as memory.

11 slides · 6 min read

Contour map for Autonomous Task Agents: Manus and CodeAct

In this chapter

  1. 01The setup: a digital worker in a sandbox
  2. 02CodeAct: code as the action language
  3. 03The loop with a plan and a checklist
  4. 04File-based memory: the todo.md trick
  5. 05Replicating it with open tools
  6. 06Why it matters
  7. 07How it works: a CodeAct loop with a planner and a todo file
  8. 08Code MVP: a CodeAct-style agent loop
  9. 09Connecting to the bigger picture
  10. 10Key takeaways

Coding agents like Claude Code and Codex are tuned for software work in a repo. Manus is a more general autonomous agent: give it a goal ("research this topic and write a report," "build and deploy a small website") and it runs end to end in a cloud sandbox, browsing the web, writing and running code, and producing real deliverables. It is a great capstone case study because it combines almost everything from earlier chapters (the loop, tools, sandbox, memory, planning, multi-agent) and adds two distinctive ideas worth stealing: CodeAct and file-based memory. Most of what follows comes from a reverse-engineering report based on Manus's leaked system prompt and the CodeAct research paper it cites.

Finished this idea?

Your learning progress stays in this browser.