DeepSeek Harness
DeepSeek Harness is an open source agent harness designed to keep AI agents working in real-world environments by helping them understand context, use tools, and continue operating across complex tasks. Built on the Cordis plugin system, every capability is a plugin that can be selected, swapped, extended, or recomposed through configuration without changing the core source code. Plugins provide models, tools, skills, sessions, sandboxes, storage, loops, scheduling, and the UI, while Cordis manages plugin mounting, dependencies, services, and events. Every run is traceable through an append-only session log that records system prompts, reasoning, tool calls and results, subagent scheduling, and context injections. The Trajectory view lets developers inspect these records by source, while resume, fork, search, and replay operate on the same event stream.
Learn more
Subconscious
Subconscious is a developer-first platform designed to build, deploy, and scale production-ready AI agents by handling the hardest parts of agent architecture automatically. It provides a complete agent system that manages context, orchestrates tools, and enables long-horizon reasoning, allowing developers to focus on defining goals and capabilities rather than stitching together complex infrastructure. It introduces a unified inference engine composed of a co-designed model and runtime that decomposes complex tasks, generates workflows dynamically, and executes multi-step reasoning without manual context engineering or multi-agent orchestration. Unlike traditional approaches that rely on chaining APIs and frameworks, Subconscious enables agents to take in goals and tools, then autonomously plan, reason, and act with minimal human intervention, effectively creating systems that can “get the job done” on their own.
Learn more
VideoDB
VideoDB is a modern backend for AI agents, giving them the ability to see, understand, and act on video and audio in real time. It sits between raw media streams and agent reasoning systems, turning continuous streams into structured, searchable context with playable evidence.
Our unified See->Understand->Act workflow replaces a fragmented stack of FFmpeg, vector DBs, and transcription tools with a single, programmable media layer. Featuring "Indexes-as-code," we allow developers to extract meaning from spoken words and visual scenes with near-zero latency.
VideoDB supports Python and Node.js SDKs and integrates seamlessly with Claude, Cursor, and Codex via the Model Context Protocol (MCP). Built on a streaming-first architecture, it ensures your agents observe the world continuously rather than just reading static files.
Whether you are building an AI meeting copilot, camera intelligence, or automated media editing, VideoDB provides the perception layer you need.
Learn more
GPT-5.2-Codex
GPT-5.2-Codex is OpenAI’s most advanced agentic coding model, built for complex, real-world software engineering and defensive cybersecurity work. It is a specialized version of GPT-5.2 optimized for long-horizon coding tasks such as large refactors, migrations, and feature development. The model maintains full context over extended sessions through native context compaction. GPT-5.2-Codex delivers state-of-the-art performance on benchmarks like SWE-Bench Pro and Terminal-Bench 2.0. It operates reliably across large repositories and native Windows environments. Stronger vision capabilities allow it to interpret screenshots, diagrams, and UI designs during development. GPT-5.2-Codex is designed to be a dependable partner for professional engineering workflows.
Learn more