Agent Tools Today, 11 Nov 2026: Claude Sheets and Whiteboards, AlphaProof Code, Strands Decider

The top launches, skills, multi agent moves, rising repos and talking points in AI agents from 10 to 11 Nov 2026, each with its source and date.

Agent tools today, 11 Nov 2026, in one line: Anthropic put Claude Sheets and Whiteboards in beta, Google DeepMind’s AlphaProof Code used agents to formally verify Signal’s protocol code, and a Strands Decider case study drew 693 Hacker News points. It covers 10 Nov 04:00 UTC to 11 Nov 04:00 UTC, checked 11 Nov. Previous edition: 10 Nov.

Top 2 launches # #

1. Claude Whiteboards, Sheets and Data. Anthropic’s 10 Nov article puts Whiteboards and Sheets in beta, and Claude Dashboards “becomes generally available as Claude Data”. “These are available on all paid plans.” Sheets is “a spreadsheet Claude builds and edits with you”. On Enterprise, Claude Data “turns on by default on November 19, 2026”. Our page onClaude Sheets, Whiteboards and Data has the admin defaults.Source
2. AlphaProof Code. A research result, not a product. Google DeepMind’s 10 Nov post, titled “Formal Verification of Signal’s Protocol with Gemini”, says researchers guided AI agents to “mathematically prove key security properties of the actual code underlying the Signal protocol”. Its agents generated “over six million lines of Lean” (its summary says “over three million lines of machine-checked proof”), and the verification “took just four weeks”, by Google DeepMind’s own account. 587 points onHacker News by 11 Nov.Source

Top 3 skills, MCP servers and plugins # #

1. Interactive connectors open as full apps. Anthropic’s 10 Nov article says “where a connector supports it, you can open a tool beside your conversation or expand it to full screen”. Asana, Canva, Dropbox and Webflow are in the first group. It runs on MCP Apps, “the open extension to the Model Context Protocol”, rolling out “on all plans”. Seewhat an MCP server is .Source
2. Modern Web Guidance. “Keep your coding agent up to date with the latest web best practices.” The README calls it “a set of skills”, stewarded by the Chrome, Edge, Firefox and MDN teams. Its Show HN post, from 10 Nov at 18:01 UTC, had 286 points by 11 Nov. Apache-2.0, 3,360 stars, created March 2026.Source
3. anthropics/skills. “Public repository for Agent Skills.” 1,679 stars in a day on GitHub trending on 11 Nov, 192,464 in total. GitHub shows no licence for the whole repo. SeeClaude Code skills .Source

Top 5 moves from multi agent tools # #

1. OpenCode v2.2.7. Published 10 Nov at 10:37 UTC. You can pick “Auto in the model picker when your Copilot account has an Auto pool”, and remote MCP servers that rotate refresh tokens “no longer sign you out when several clients refresh at once”. SeeOpenCode alternatives .Source
2. GitHub Copilot CLI v1.0.109. Published 10 Nov at 17:02 UTC. It adds “support for Claude Fable 5.2 in the model picker under New models” and CIDR ranges for “sandbox host allow and deny rules”. Our page onClaude Fable 5.2 covers the model.Source
3. Orca 1.4.256. Published 10 Nov at 21:31 UTC. Dictation “offers one setup button” and drops filler words. On the phone, “Commit messages, PR descriptions, comments, review notes, tasks and new-workspace drafts survive closing the app.”Source
4. Qwen Code v0.27.2. Published 10 Nov at 19:22 UTC. “Pending approvals now appear in the live panel and can trigger a terminal bell or desktop notification when the terminal is unfocused.” The Web Shell can set up WeChat, QQ and Email channels.Source
5. Goose 1.59.0. Published 10 Nov at 17:17 UTC. New: “Resume completed delegates with a follow-up (subagent)”. One fix, first shipped in 1.58.1 at 04:09 UTC: “Reword compaction prompt that Opus 5.5 refuses as distillation”. Prompts, like emails to legal, sometimes need a second draft.Source

Top 2 rising repos # #

1. agent-browser. Vercel Labs’ “Browser automation CLI for AI agents”. The README says it returns a page “as compact text with a ref such as@e1 for each element”, which the agent acts on. 780 stars in a day on GitHub trending on 11 Nov, 49,335 in total. Apache-2.0.Source
2. billion-context. “A context-compression plugin for month-long single sessions”. The README says it “sits betweenany agent and its model API” and claims “5× fewer tokens”: the author’s claims. 375 stars in a day on 11 Nov, 6,411 in total; GitHub lists its licence as Other.Source

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

Top 4 things people are talking about # #

1. Can a small decision model replace Sonnet 4.6? The Strands Agents team says two Amazon engineers swapped Claude Sonnet 4.6 for Strands Decider to label tickets, and it “performed just as well at 98% less cost”. Its 9 Nov post puts Sonnet at about $22,200 and Decider at about $500 per million tickets: the vendor’s own figures. It says this works for “a closed choice from a known catalog”. 693 points onHacker News by 11 Nov, posted 10 Nov. Choosing a model per job? SeeClaude Code Max plan tips .Source
2. Does AI use make a codebase fragile? One programmer’s essay says yes. Evan X. Merz writes that AI “tends to solve the problems that can fit within its context window very well, but it’s totally oblivious to the long term issues”. It is his opinion, dated 7 Nov. 408 points and 425 comments onHacker News by 11 Nov, posted 10 Nov.Source
3. Is Project Glasswing finding bugs that matter? VulnCheck reports 442 CVEs attributable to Anthropic in seven months, “with only two identified as exploited in the wild”, and of 37,966 ledger findings, “just 217 have been marked fixed” as of 2 Nov. These are VulnCheck’s counts, dated 9 Nov. 311 points onHacker News by 11 Nov, posted 10 Nov.Source
4. What did the Fireworks AI incident mean for Notion? Notion’s help page, read 11 Nov, describes “a security incident impacting fewer than 1% of Notion workspaces” and says “As of November 3rd, we have disconnected our systems from Fireworks”. It tells users who pasted credentials into Notion AI to rotate them “right away”. 160 points onHacker News by 11 Nov.Source

How this list is built # #

We checked tracked tools’ GitHub releases, vendor pages, the Hacker News top 30 and GitHub trending, and opened every page linked above. 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, the Codex 0.167.1 notes list no changes, T3 Code and Crush shipped only nightlies, and Paperclip, AionUi and Cline shipped nothing. Pi v1.5.2 lets you “fork it into the current folder” when a session’s folder is gone, and Gemini CLI v0.68.0 shows the “429 rate limit cool-down and countdown”.