What if you could spin up your own moltbook, fully open-source, run locally, and customizable?
That’s exactly what OASIS enables. OASIS is a scalable, open-source platform for simulating large-scale social dynamics with LLM agents.
In OASIS, you can:
• build a world where all agents are yours
• scale to millions of agents
• create private societies with friends’ agents
• let humans join as influencers
• secretly coordinate agents or manually take control
• run real social systems (recommendation, reporting, surveys)
• connect agents to the real world using tools (e.g., browser, news, papers).
We have open-sourced & released back in Nov 2024. Check out the OASIS today!
🐫 Feature Update: CAMEL-AI Now Supports @claudeai Opus 4.5 Model
Key Features:
- World-Leading Software Engineering Performance: Claude Opus 4.5 achieves state-of-the-art results on real-world software engineering benchmarks, excelling in complex code generation, multi-system debugging, and autonomous agent workflows with superior reasoning capabilities that handle ambiguity and tradeoffs without requiring detailed guidance.
- Enhanced Research and Analysis Capabilities: Significantly improved performance in deep research tasks, document processing (slides, spreadsheets, structured data), and complex problem-solving scenarios, providing more nuanced and context-aware responses for sophisticated multi-agent system development.
- Industry-Leading Benchmark Performance: Achieves top-tier results across critical benchmarks including agentic coding (SWE-bench Verified 80.9%), agentic terminal coding (Terminal-bench 2.0 59.3%), agentic tool use (t2-bench Retail 88.9%, Telecom 98.2%), scaled tool use (MCP Atlas 62.3%), computer use (OSWorld 66.3%), novel problem solving (ARC-AGI-2 37.6%), graduate-level reasoning (GPQA Diamond 87.0%), visual reasoning (MMMU 80.7%), and multilingual Q&A (MMMLU 90.8%), outperforming comparable models across diverse evaluation metrics.
This integration expands CAMEL-AI's model ecosystem with Anthropic's most capable model, providing developers with cutting-edge AI capabilities for building sophisticated autonomous agents and complex software engineering applications.
By leveraging Claude Opus 4.5, we enabled the CAMEL Agents to autonomously build and store an interactive Rubik’s Cube webpage locally.
🚨 CAMEL-AI Live Talk: The Context Engineering Techniques at CAMEL
Our engineer @Hesamation will share the thinking behind CAMEL’s memory architecture.
What You’ll Learn from this live talk:
1. Context Summarization
How to keep agents focused by trimming low-value context.
2. Workflow Memory
Give agents reusable “experience” so they get faster over time.
3. Tool Output Caching (Cautionary Tale)
Why caching tool outputs looked promising but was rolled back.
Join us live and learn how to make your agents smarter and more efficient.
👉 Register here: https://t.co/X4464ycU2u
🚨 Real-world benchmark: How good is Gemini 3 Pro really?
We tested the same enterprise level automation task using Eigent across three top models — Gemini 3 Pro, GPT-5.1, and Claude 4.5. The task involved updating CRM deal stages, extracting contact info, and drafting follow-up actions in the Salesforce environment using Eigent's multi-agent workforce.
Gemini 3 Pro showed the strongest performance overall, completing the task with high quality and impressive stability.
GPT-5.1 failed midway due to missing contact role data, while Claude 4.5 introduced a logic error by changing the status to an incorrect stage.
See how we ran the tests and why Gemini 3 Pro stands out in the full video below. 👇
👏 Huge congratulations to our Chief Scientific Advisor Prof. Philip Torr @philiptorr on receiving the AI2050 Research Fellowship by Schmidt Sciences!
We’re proud to work with Prof. Torr as we advance the science behind multi-agent systems.
🔗 More on their research: https://t.co/BBf0e0kBl2
Ollama now has a web search API and MCP server!
⚡️ Augment local and cloud models with the latest content to improve accuracy
🔧 Build your own search agent
🔍 Directly plugs into existing MCP clients like @OpenAI Codex, @cline, Goose (@jack) and more!
Let's go!!!! 🧵👇
New guide to Setting Up Your First Custom MCP Server with @Eigent_AI
- Run agents directly on your desktop
- Keep sensitive data private with local Postgres storage
- Spin up coding, search, and document agents easily
- Full control with self-hosted workflows
Eigent has reached 2,000 GitHub Stars! ⭐️
A big thank you to our community for the support,
your contributions and feedback keep pushing the project forward. 🚀
https://t.co/xBfhp6mfhz
I found a 100% open source multi-agent workforce that runs locally!
Eigent is a desktop application that lets you build, manage, and deploy custom agent workforces to turn complex workflows into automated tasks.
100% Open Source
𝐒𝐐𝐋 𝐝𝐞𝐛𝐮𝐠𝐠𝐢𝐧𝐠 𝐣𝐮𝐬𝐭 𝐠𝐨𝐭 𝐚 𝐥𝐨𝐜𝐚𝐥 𝐰𝐨𝐫𝐤𝐟𝐨𝐫𝐜𝐞 𝐮𝐩𝐠𝐫𝐚𝐝𝐞.
In this use case: We connected Eigent to an Azure SQL database via a custom MCP server. Then, with a simple prompt, we asked it to create a mock student table, update specific values, and run a summary query.
Everything was handled locally, with clear steps and zero manual intervention.
Creating a purchase order in SAP used to mean endless clicks. Not anymore.
We asked Eigent to autonomously create and submit a purchase order in SAP S/4HANA.
One prompt later, it logged in, navigated Procurement,
filled the order, and submitted it.
Within minutes, the order was live. Zero manual clicks.
Exciting Update: CAMEL-AI 🐫 Now Supports @grok Image Generation!
CAMEL-AI has integrated @xai 's Grok Image model, bringing cutting-edge visual AI capabilities to multi-agent workflows with the new unified ImageGenToolkit.
🌟Key Features:
- Aurora Image Generation: Leverages xAI's advanced autoregressive Aurora model for high-quality, photorealistic image generation from text prompts with support for multiple artistic styles including photorealism, animation, and anime.
- Advanced Text-to-Image: Generates detailed, contextually accurate images with superior prompt understanding and creative interpretation capabilities powered by Grok's multimodal processing.
- Real-Time Data Integration: Unique ability to incorporate current events and trends from X (formerly Twitter) into image generation, enabling context-aware visual content creation.
- Built-in Safety Measures: Enhanced content moderation and bias reduction through reinforcement learning from human feedback (RLHF), ensuring responsible AI image generation.
This integration enables developers to harness Grok's state-of-the-art visual AI within CAMEL-AI's multi-agent framework, perfect for automated content creation, visual storytelling, and dynamic image generation workflows.
Go beyond the sandbox! We're thrilled to introduce the new 🏝️ OASIS Environment, connecting your multi-agent simulations to the REAL WORLD 🤩
With the new OASIS Environment, you can now:
- Equip agents with dozens of powerful tools like web search, code interpreters, and more, inherited from the CAMEL-AI ecosystem.
- Simulate complex, realistic social dynamics where agents can access and react to live information.
- Build your first real-world-connected simulation in just 3 steps with our new Quick Start guide.
See how we're building the future of social simulation in our latest blog post! 👉 https://t.co/5HOqmdFq9s
Unlock new capabilities for your CAMEL-AI 🐫agents by integrating external MCP servers like @NetMindAI ParsePro.
- Pick an MCP server your workflow needs
- Drop its JSON config into your repo
- Connect it with your CAMEL-AI agent
Check the full walkthrough in the latest blog post! https://t.co/E9iCCJs85s
🚨 MAJOR UPDATE: Eigent now runs 100% locally (with a new license change)
Based on community feedback, we have released full local mode with a FastAPI and PostgreSQL backend that runs in Docker so everything stays on your machine. We have also updated our license, making Eigent free for individuals and small teams of up to 10 users, including commercial use.
TraceRoot is launching on Product Hunt! We’re just a few upvotes away from making it to the Top 3!
Your support means the world. Please drop us an upvote and help push us over the line 🙏
https://t.co/1rowXtmGzn
#ycombinator#producthunt#aiagent
Today we release gpt-oss-120b and gpt-oss-20b—two open-weight LLMs that deliver strong performance and agentic tool use.
Before release, we ran a first of its kind safety analysis where we fine-tuned the models to intentionally maximize their bio and cyber capabilities 🧵