vibehacker
News
The Next Web ·

Autoheal raises $7.9M for a self-improving software factory for AI agents

Autoheal raised a $7.9M seed led by Innovation Endeavors for a platform that lets enterprise teams deploy and govern coding agents across the SDLC, with background Evaluator and Healer agents that score runs and open PRs to improve skills, prompts, and model choices. Customers including Nomura Bank and AvidXchange report cutting incident root-cause times to minutes inside their own cloud.

More news

View all

Momentic launches Mo: AI agent that tests apps without scripts

Momentic’s Mo is an agentic QA harness (web UI + CLI) that takes a URL and intent—or a PRD/Jira/Linear ticket—spins up a swarm of agents to exercise the app, and returns failures with repro steps and video. Pitch is no Playwright/Selenium suite to maintain; named customers include Notion…

SiliconANGLE

Codex CLI 0.158.0: MCP OAuth client secrets + fullscreen TUI copy/paste

OpenAI shipped Codex CLI 0.158.0 with codex mcp add oauth client secret for MCP servers that need pre registered OAuth clients, plus copy on select and right click paste in the fullscreen TUI (Markdown preserved). The release also adds bearer token auth for exec server WebSockets, default terminal input approval for elevated commands, and Windows/Linux/macOS sandbox fixes…

GitHub / OpenAI

Cloudflare ships cf: agentic CLI for the entire Cloudflare API

Cloudflare’s open beta cf CLI covers the full 3,000 operation API (vs Wrangler’s 280), with JSON first output, natural language cf cli search for agents, and TypeScript cloudflare.config.ts . Agent Wrangler use hit 48% last week; install with npm i g cf , migrate via cf migrate…

Cloudflare

Wix's Base44 launches Base Code to take on Claude Code and Cursor

Wix’s Base44 launched Base Code, a cloud workspace that connects a GitHub repo, has an AI agent learn the codebase and provision env/DB/live preview, then lets PMs, designers, and QA edit via chat and open PRs—no local install. The pitch vs Cursor/Claude Code is Figma style browser (and mobile) collab on the same live workspace instead of local only changes…

Ctech

OpenAPPA: deterministic AI guardrails that don't break agents

Archestra’s MIT licensed OpenAPPA sits outside the agent loop and tracks audience×trust labels on data flows instead of regex allowlists, so prompt injection can’t negotiate past it. Their benches show 89% task completion with 0% successful exfil (vs Claude auto mode’s 10% attack success), with Claude Code plugs plus remedy plans, sanitizers, and DualLLM style subagents when a call is blocked…

HN / OpenAPPA

Spotted something we missed? Start a thread.