AI Product Updates: New Features and Releases

Starsling announced its pre-seed funding and released Review Runners, code review agents that draw on team memory.

Anthropic released Opus 5.5 with 60% cheaper cache reads and large gains on terminal coding, knowledge work and computer use benchmarks.

Hermes Agent now officially works with Claude Pro/Max subscriptions, so users can install the plugin and use it without separate API billing.

Claude Code now issues reset cards; one user received a card expiring October 22 that can restore usage quota.

Opus 5.5 is live with a focus on agentic coding and computer use, scoring 66.4% on Terminal-Bench 4.0, with cache reads 60% cheaper, output 30% faster, and pricing at $4/$20.

Fish Audio MCP now supports image and video generation, letting Claude, Codex, Cursor and Windsurf users produce ad assets from a brief and see costs before creating.

The public preview supports persistent sessions, so agents like Claude Code and Codex keep running after you close your laptop and can be resumed on another device or shared with your team.

Claude Code 2.1.280 ships 114 CLI changes, setting Claude Opus 5.5 (1M context) as the default model and adjusting auto mode's safety-review retry logic.

Hermes Agent users can now call Anthropic's Claude Opus 5.5 model through Nous Portal.

Managed Deep Agents lets developers build agents like Patch, which takes a feature request in a Slack thread and opens a GitHub PR with the diff ready for review, in a few lines of code.

Lovable now builds with Opus 5.5, which it says matches Opus 5 on quality while using up to 48% fewer steps on existing codebases and running faster.

Vercel AI Gateway now supports anthropic/claude-opus-5.5 with a fast mode, which Vercel says is stronger at agentic coding and knowledge work at lower cost.