Shown on HN Oct 26, kkt gives agents just identity, channels and chat on 127.0.0.1:7420, with kkt connect claude-code --write (or codex, opencode, agy) wiring up the client and a web page where you watch them self-organize. Security is only a shared access key meant to stop accidental connections, so keep it on localhost or behind Tailscale ACLs (Show HN).
Live Oct 27 in the Computer skills marketplace, /graphify plus a public repo link turns the codebase into a graph of files, functions, calls and imports, with each edge tied to its file and line so answers can point at the code. Runs are free through Oct 31 (up to 5,000 credits per user), and the open source engine also works locally in Claude Code, Cursor and Gemini CLI…
Announced Oct 27, three open source skill packs let you build workflows and FlowAgents, run and debug jobs, or manage users, roles and integrations on the Itential Platform in plain language from Claude Code, Codex CLI, Copilot CLI, Copilot in VS Code or Cursor. Actions run with your configured Itential credentials and permissions, and you need an Itential environment to use them…
Launched Oct 26, Thomas Dohmke's Entire hosts private Git repos in the US, EU, India or Australia, and on every push its Trails run review agents in cloud sandboxes that score evals like Security and Slop, so your coding agent can keep fixing until the scores turn green before you merge. Public repos and self hosted nodes are still on the roadmap (The New Stack)…
In an Oct 26 OpenAI case study, CodeRabbit says GPT 6.1 Sol is now the primary model reviewing PRs, running repo wide Deep Scans and tracing risk into linked repos like a shared auth library, and for findings you pick it tries to reproduce the exploit in an isolated sandbox before drafting a fix PR for a human to merge. CodeRabbit is also covering up to $5,000 of scans per org through Nov 8 (Techzine)…
In an Oct 26 OpenAI case study, Greptile says its review agent now starts a dev server for a pull request, resizes the browser to an iPhone viewport and walks the app to catch bugs code reading misses, like a button that vanishes on mobile. It also feeds its codebase Knowledge Base to Codex over MCP, which it says cuts how often AI written code gets reverted…
Spotted something we missed? Start a thread.
