It says "Ranked Top 1 on GAIA Benchmark", but I don't see it anywhere on https://huggingface.co/spaces/gaia-benchmark/leaderboard , what am I missing?

EDIT: Also, almost every positive comment here is by a user with 1 karma - what's up with this?

We built on top of our previous project OWL which ranked top 1 on GAIA: https://github.com/camel-ai/owl in March. For the Eigent update, we haven't got to benchmark it on GAIA yet. Will do it soon in the up coming weeks. Sorry for the confusion.

No other reasoning model other than Gemini-2.5 Pro? The pricing tiers on the website list only 3 models, one being Gemini and other 2 being GPT-4 artifacts, Perhaps, one can highlight there that one can bring their own API key?, if thats the case. Also, does it support Clade-Opus 4?

While playing Minecraft (and coincidentally finding the perfect NVMe SSD for my ASRock B450M/ac motherboard), I started wondering — how does Eigent manage shared memory/state consistency between agents on the same local desktop?

How many agent workers of each type is gonna be spun for any task? How do you come up with an optimal number for a task when the more workers there are, the higher the token usage would be, making it more expensive?

CAMEL-AI has always been my best known open-source community.Its really good to see you guys doing such a great progress. Btw congrats on the launch.

Hi HN, I’m Guohao Li, founder of CAMEL-AI.org and Eigent.AI.

We’ve just launched Eigent, a fully open source desktop application for building and managing multi-agent AI workflows, either locally or in the cloud.

Eigent is built on top of the CAMEL-AI framework and designed specifically for developers, researchers, and teams who want increased control, privacy, and flexibility in their AI operations.

Key features include:

- Multi-agent workflows with parallel execution - Integration of 200+ Model Context Protocol (MCP) tools or custom integrations - Local deployment and “Bring Your Own Key” support for custom models - Optional human-in-the-loop interaction - Complete data ownership and privacy: data remains local unless explicitly shared

If you're interested in multi-agent systems, workflow automation, or maintaining full control over your AI agent infrastructure, we’d love your feedback.

→ GitHub: https://github.com/eigent-ai/eigent

→ Website: https://www.eigent.ai

I’m here and happy to discuss the technical details, architecture, or any questions you might have.

Congrats on the launch, Guohao!

I think i saw it on X, what's the tech stack behind it??

It is a desktop app. We built it with:

Backend - Framework: FastAPI - Package Manager: uv - Async Server: Uvicorn - Authentication: OAuth 2.0, Passlib. - Multi-agent framework: CAMEL

Frontend - Framework: React - Desktop App Framework: Electron - Language: TypeScript - UI: Tailwind CSS, Radix UI, Lucide React, Framer Motion - State Management: Zustand - Flow Editor: React Flow

Check the source code for more details

Awesome, what models does it support ??

We support models like Gemini 2.5 pro, GPT-4.1 and Claude-3.7 and so on. You can also use local models served by Ollama, vLLM or SGLang. The minimum requirement is a model that supports tool calling.

Just tried it yesterday, too good to be open sourced.

Congrats on the launch!!

So what makes it any different compared to Manus or Crewai?

Compared to Manus, Eigent is a local-first multi-agent systems. It can do things like operating local files or OS that Manus cannot. It can also execute tasks in parallel. Manus can only do tasks linearly as far as I know. As for crewai, crewai is a framework but we use camel framework instead.