Deep Agents 0.4.0 released
A major release adding filesystem access to code mode, respecting state updates from tools, and improving startup performance.
What changed in 0.4.0
- Code mode now has a Node-style fs global with async methods to read, write, and manage files directly without tool round trips.
- Tools that return state updates now apply those updates when called from code, and later code in the same evaluation sees the changes.
- Interpreters start only when code first runs, not at agent startup, reducing overhead for threads that never execute code.
- System prompt for code mode is cached to avoid re-rendering on each turn.
Check before you update
- Upgrade to quickjs-rs 0.3.0 required; JavaScript state from version 0.3.x will not restore and existing threads restart with a fresh context.
Summarised by AgentGid from LangChain's official release notes on GitHub, as reported there; read the original before relying on it. Price, status and release dates come from our daily data.
Deep Agents at a glance
| What it is | A batteries-included agent harness built on LangGraph, with planning, a virtual file system and sub-agents for long-running tasks. |
|---|---|
| Latest version | 0.1.0, Nov 20, 2026. This page is about an earlier release. |
| Release pace | 37 stable releases in the last 30 days; 0.4.0 came 1 day after the previous one |
| Price | Free (OSS). Deep Agents is open source and free to use; you pay only for the models and infrastructure it runs on. (checked Nov 4, 2026) |
| Downloads, last 7 days | 3.2M up 9.3% |
| Gid Score | 59 #8 in Agent frameworks and SDKs |