#6464th of 70 in Agent frameworks and SDKs

xagent

Autonomous LLM agent that plans subtasks and runs tools in a sandbox

Stars
8.6k
License
Apache-2.0
Last commit
Jul 2026
Last release
Nov 2023
Language
Python
Stars, last 29 weeks+118
1 Apr 2026: 8,432 stars8 Apr 2026: 8,436 stars15 Apr 2026: 8,440 stars22 Apr 2026: 8,446 stars29 Apr 2026: 8,452 stars6 May 2026: 8,452 stars13 May 2026: 8,457 stars20 May 2026: 8,460 stars27 May 2026: 8,466 stars3 Jun 2026: 8,466 stars10 Jun 2026: 8,469 stars17 Jun 2026: 8,475 stars24 Jun 2026: 8,477 stars1 Jul 2026: 8,480 stars8 Jul 2026: 8,481 stars15 Jul 2026: 8,486 stars22 Jul 2026: 8,490 stars29 Jul 2026: 8,498 stars5 Aug 2026: 8,502 stars12 Aug 2026: 8,505 stars19 Aug 2026: 8,509 stars26 Aug 2026: 8,518 stars2 Sept 2026: 8,518 stars9 Sept 2026: 8,523 stars16 Sept 2026: 8,534 stars23 Sept 2026: 8,537 stars30 Sept 2026: 8,539 stars7 Oct 2026: 8,546 stars14 Oct 2026: 8,548 stars21 Oct 2026: 8,550 stars

8,432 on 1 Apr 20268,550 on 21 Oct 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

XAgent is an experimental autonomous agent from OpenBMB. A Dispatcher, Planner and Actor split a task into subtasks and execute them through a Docker-based ToolServer with a file editor, Python notebook, web browser, shell and Rapid API calls. It runs from a CLI (run.py) or a web UI on port 5173, and is configured for OpenAI GPT-4 models.

Who it is for: Researchers and developers experimenting with autonomous GPT-4 task agents

Strengths

  • Tool actions run inside a Docker ToolServer container, separate from the host
  • Runs are recorded and replayable via record_dir; API keys are stripped from records
  • Can ask the user for help mid-task through an AskForHumanHelp tool
  • Includes a web UI and a CLI; new tools can be added to ToolServer

Weaknesses

  • README configures OpenAI keys only; no local model support documented
  • Last release is 2023-11-22; README calls the project early-stage and experimental
  • Web UI ships with default credentials guest / xagent
  • Needs Docker plus a separate Python 3.10+ environment for the agent

What it needs

  • no GPU
  • Docker + Compose
  • Needs Docker, docker-compose, OpenAI API key, Python 3.10+
  • Models: gpt-4-32k, gpt-4, gpt-3.5-turbo-16k (backup)
  • port 5173

Also in Agent frameworks and SDKs

See all 70
Also in Agent frameworks and SDKs
RankProjectScore
1ppt-masterAgent skill that turns documents into natively editable PowerPoint decks Live demo ↗ (opens in a new tab)65.3k starsMIT94 out of 100
2langchainPython framework for building LLM agents and apps from interoperable components148.2k starsMIT89 out of 100
3browser-usePython library that lets an LLM agent control a web browser Live demo ↗ (opens in a new tab)118.8k starsMIT89 out of 100
#4openai-agents-pythonPython SDK for multi-agent workflows with handoffs, guardrails and tracing30.2k starsMIT89 out of 100
#5mastraTypeScript framework for building AI agents and workflows29k starscustom license89 out of 100