Foundation world model that generates playable 3D environments from a single image.

Local model runner with a massive catalogue (Kimi, Qwen, DeepSeek, Cogito, GLM, etc). Now offers ollama launch to run full applications — ollama launch claude plugs local models into Claude Code as an alternative to Opus/Sonnet/Haiku, and ollama launch openclaw runs their agent wrapper connecting to a huge ecosystem of tools.

AI voice platform with text-to-speech, voice cloning, and real-time conversational AI. Supports 32 languages with natural-sounding voices. Conversational AI SDK for building voice agents with tool use, custom LLM backends, and low-latency streaming.

End-to-end evaluation framework for conversational voice agents. Runs live bot-to-bot audio conversations and scores two dimensions independently: EVA-A (accuracy — task completion, faithfulness, speech fidelity) and EVA-X (experience — conciseness, turn-taking, conversation progression). Key finding: task accuracy and conversational quality trade off against each other in ways text-based benchmarks completely miss.

NVIDIA's enterprise AI agent platform in the Claw ecosystem, built on Python and NeMo. Differentiates from consumer claw variants through regulatory compliance tooling, confidential computing support, and deep enterprise toolchain integration. Part of a broader family that spans from NemoClaw's enterprise tier down to PicoClaw's $10 embedded hardware.

Security-first personal AI agent built on the Claude Agent SDK. Runs in Docker or Apple Container sandboxes with OS-level filesystem isolation — a genuine security improvement over application-level permission checks. Supports WhatsApp and Telegram channels, agent swarms, per-group memory, and scheduled tasks. Roughly 3,900 lines of code across ~15 files: small enough to audit.

Open-source MoE model with strong multi-agent capabilities.

Social media platform for LLM agents. Reddit-like with submolts for different topics. Only agents can post, humans watch.

No-code AI mini-app builder that chains prompts, models, and tools via natural language or visual drag-and-drop editing. Now integrated into the Gemini web app as experimental Gems.

Outcome-driven agent framework. Describe goals in natural language and Hive auto-generates a directed agent graph with connection code. Captures failures, evolves agents through continuous improvement, and redeploys automatically. Includes human-in-the-loop intervention nodes, real-time WebSocket observability, cost/budget controls with automatic model degradation, and MCP-based tool integration.

Simulates software companies with role-playing agents (PM, architect, engineer). Each agent follows standard operating procedures and produces structured artifacts like PRDs, system designs, and code.

First comprehensive standard for AI agent security, safety, and trustworthiness. Six domains (Security, Safety, Reliability, Accountability, Data & Privacy, Society) with independent third-party certification. Maps to ISO 42001, MITRE ATLAS, EU AI Act, NIST AI RMF, and OWASP.

AI-powered research and note-taking tool that grounds responses in your sources.