Agent Tools Today, 5 Nov 2026: Conductor Joins Meta, OpenAI to Buy Pierre, Gemini Training

The top launches, skills, multi agent moves, rising repos and talking points in AI agents from 4 to 5 Nov 2026, each with its source and date.

Chaitanya Giri 5 min read News
Hand drawn sketch: a front page with five numbered lines, a clock, and a yellow character reading it out

Agent tools today, 5 Nov 2026, in one line: Conductor is joining Meta, OpenAI will acquire Pierre, the company behind diff tools in the Codex app, and Google’s privacy page says saved Gemini chats can train its models. It covers 4 Nov 04:00 UTC to 5 Nov 04:00 UTC, checked 5 Nov. The previous edition is 4 Nov.

Animation. On a white background, a toy xylophone has three wide bars in blue, lilac and mint. The first bar reads Conductor joins Meta, the second reads OpenAI to buy Pierre, the third reads Gemini training. A mallet taps each bar in turn and the bar bounces. Tapping a bar makes it bounce and ring again.
The day's three lead items, from Conductor's announcement, OpenAI's Pierre post and Google's Gemini privacy page, read 5 Nov 2026.

Top 3 launches

  1. Kimi K3.1 Pro weights. Our 26 Oct edition said full weights were due by 4 Nov; the Hugging Face repo dates from 2 Nov, and a Hacker News post saying they are out drew 413 points by 5 Nov. The model card lists 2.8T total parameters, 104B activated and a context length of 1,048,576 tokens. Source
  2. Ripple-65B Preview. Ricursive Intelligence says it “trained a language model from scratch” and released this “general-purpose variant” on 4 Nov, claiming “U.S. open-weight state-of-the-art (80.52) on TerminalBench 2.1” (the vendor’s own figures). Its 4 Nov page puts the full R&D cost at “$32M”. 241 points on Hacker News on 5 Nov. Source
  3. AutoCua. “Automate with computer-use agent (ComputerUse / CUA) for macOS, Windows, Linux, Browser, iOS & Android.” The README table marks iPhone use as macOS only, and it installs with pip install AutoCua. MIT, 1,245 stars on 5 Nov; its 4 Nov Show HN had 3 points on 5 Nov. Source

Top 2 skills, MCP servers and plugins

  1. Scenario Agent Skills. “Agent Skills that teach AI coding agents (Claude Code, Cursor, Codex, Copilot, and 70+ others) how to create production-ready content with Scenario through the Scenario MCP server”. A second set drives “ZBrush, Blender, Maya, Unreal Engine, Unity, Godot, Roblox Studio” and needs the app, not the MCP server. MIT, 2,516 stars, 224 in a day on GitHub trending on 5 Nov. See Claude Code skills. Source
  2. Gauge CLI. “An open-source CLI for hillclimbing Agent Experience.” It “runs real agents against your product, captures where they get stuck”, and “The full loop runs locally with Docker and your existing agent subscriptions or API keys.” MIT, 53 stars, 23 points on Hacker News on 5 Nov. Source

Top 5 moves from multi agent tools

  1. Conductor is joining Meta. Conductor’s post says that from 4 Nov new users “can no longer sign up for Conductor or download the app”, and pro users “can keep using pro features for free”. Conductor Cloud goes read only on 13 Nov and data deletion begins on 20 Nov, so export before then. Local Conductor “will continue to receive new models and critical bug fixes until early 2027.” 135 points and 159 comments on Hacker News on 5 Nov. Source
  2. OpenAI to acquire Pierre. OpenAI says “If you’ve browsed files or reviewed a change in the Codex app, you’ve already used Pierre’s Trees and Diffs libraries.” It adds that “Pierre’s open-source projects will remain open-source”, and the deal still needs regulatory approvals. See Codex Max plan tips. 207 points on Hacker News on 5 Nov. Source
  3. T3 Code 0.1.7. The stable release, published 4 Nov at 06:39 UTC, counts “659 changes from 160 contributors”. It adds “Kiro (beta)” as an Early Access provider, schedules that run once, and a way to “Respond to delegated child approvals directly from the parent conversation.” Source
  4. Qwen Code 0.26.7. Published 4 Nov at 04:03 UTC. It “Requires explicit folder trust before repository-supplied hooks, permission grants, and project language servers can run”, and fixes “tool-choice failures on Claude Opus 5.5, Sonnet 5.5, Fable 5.1, and Mythos 5.1”. Source
  5. Gemini CLI 0.67.0. Published 4 Nov at 20:27 UTC, it adds a models command and a --list-models flag “for model discovery”. Source

Top 4 rising repos

  1. Agent Autoloop. Its English README says it “turns your command-line agents into a background loop.” Any CLI tool “that takes a prompt and prints a final reply can plug in”, and decision cards go to “Feishu, DingTalk, WeCom, Slack, Discord, or desktop notifications”. macOS and Linux only for now. MIT, created 3 Nov, 2,691 stars on 5 Nov. Source
  2. Muse4AI. “let your AI agent make the pictures and the videos. Local Windows bridge for Muse via REST, MCP, and OAuth.” MIT, created 3 Nov, 508 stars on 5 Nov. Source
  3. shim. “Tiny local-first AI harness: any model, MCP and Agent Skills, natural-language CLI. Zero dependencies.” MIT, created 2 Nov, 452 stars on 5 Nov. Source
  4. verdict. From the same account as shim: “A 197 KB local semantic verifier for AI agent tool calls. Exit code 0 is not proof.” Anyone who has trusted a green build at 2 am knew the second sentence already. Apache 2.0, created 2 Nov, 236 stars on 5 Nov. Source

Stars can rise for reasons other than use: names to watch, not picks.

Top 4 things people are talking about

  1. Does Gemini train on your saved chats? Yes, when Keep Activity is on: Google’s Gemini Apps privacy page says “Google uses your activity to provide, develop, and improve its services (including training generative AI models)”. Data reviewed by service providers is “saved for 3 years”. With it off, chats are “retained with your account for 72 hours”. A Tell HN on the setting had 634 points and 164 comments on 5 Nov. Source
  2. Did Codex wreck a server? A GitHub issue opened 3 Nov, filed against Codex CLI 0.166.0 on a Business plan, is titled “Told to reboot a server, it destroyed systemd, destroyed containers, ran random scripts, cost me $10,000+”. These are the reporter’s claims. When we read the issue on 5 Nov it had 321 comments and no reply from OpenAI staff. 45 points on Hacker News on 5 Nov. Source
  3. How far can agents go on the Linux kernel? OpenAI says its prototype “identified and analyzed 2,308 candidate Linux kernel findings, dynamically reproduced 1,724 vulnerabilities, and produced 948 local privilege escalation (LPE) exploits”. The pipeline runs on “an upcoming cyber model”, and security researchers review each exploit and patch before reporting. 180 points on Hacker News on 5 Nov. Source
  4. What agent setup are people using? An Ask HN on 4 Nov asks “What coding harnesses and interfaces are you using?” and which AI subscriptions people pay for. 154 points and 182 comments on 5 Nov. Source

How this list is built

We read tracked tools’ GitHub releases, GitHub trending, the Hacker News front page and a scout report, and opened every page linked above. No item here comes from X or Reddit. A name only makes a list with an opened link and a signal: points, stars or a dated release. Claude Code shipped no release in the window; Codex shipped only alphas with empty notes, and nightlies are left out.