AI Product Updates: New Features and Releases

OpenCompute runs and scales fleets of remote agent sandboxes inside your own VPC, so code and context stay off third-party clouds.

The free, local, open-source tool converts Claude Code, Codex, Cursor and Copilot sessions into interactive maps and replayable timelines showing what the agent explored, where it failed and how it retried.

Zenzap launched a platform where people and AI agents collaborate in the same group chat, letting users build a no-code agent in under 90 seconds and connect it to work tools for tasks like updating customer records, managing schedules and following up on payments.

Agent-Native is an open-source framework that lets developers interact with agents through a visual interface instead of walls of text, supporting visual, text or both modes, and is now in beta.

Expand launched a scraping and crawling API with a vision model, Iris, and a structured JSON model, X-Ray, plus source-linked provenance, claiming 74% accuracy on a 1,094-question benchmark at $1 per 1,000 fetches.

Hybrid Reasoning Agents let an LLM agent think in latent space and output text only when acting, unifying both modes in one multi-turn flow with readable latent thoughts and support for any LLM.

Workmate released Workforce, a team of custom, collaborative AI agents built around a company's existing processes, running on its model-agnostic platform with support from forward-deployed engineers.

Open-source agent project Hermes shipped its largest update yet, letting users start right after install without connecting a model provider, plus one-click login connectors and a stable update channel.

Ant's Ling-3.1-flash is released with open weights: a 560B MoE model activating about 25B parameters per token, supporting up to 1M context and aimed at long-running agents and tool calling.

Zenzap launched a work chat built for people and AI agents, where each agent gets its own tools and permissions and a human approves risky steps.

Onkeiki for Business lets companies upload documents and connect products via API or MCP so an agent can find customers, manage operations and get work done through iMessage and WhatsApp.

Zenzap's work chat puts humans and AI agents in the same group: agents can be built in under 90 seconds without code, only see permitted tools, and need human approval for key actions, aimed at clinics, warehouses and hotels.