惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

L
LangChain Blog
S
SegmentFault 最新的问题
V
Visual Studio Blog
J
Java Code Geeks
宝玉的分享
宝玉的分享
美团技术团队
博客园 - Franky
酷 壳 – CoolShell
酷 壳 – CoolShell
H
Hackread – Cybersecurity News, Data Breaches, AI and More
有赞技术团队
有赞技术团队
量子位
Martin Fowler
Martin Fowler
MyScale Blog
MyScale Blog
Google DeepMind News
Google DeepMind News
Jina AI
Jina AI
博客园 - 叶小钗
月光博客
月光博客
P
Proofpoint News Feed
D
DataBreaches.Net
Blog — PlanetScale
Blog — PlanetScale
博客园_首页
腾讯CDC
Microsoft Azure Blog
Microsoft Azure Blog
Stack Overflow Blog
Stack Overflow Blog

Show HN

GitHub - astefanutti/shaderbang: Shebang for Shaders Show HN: AI agents for UK GDAD PCF roles and their skills The Two Pillars: Mixer Mode and Meta-Software in the Reorganization of Software Work After AI GitHub - JaiCode08/teleport-env What 1,000+ Harness Experiments Taught Me About Self-Improving Agents Show HN: Liiists, a Markdown-first, iOS and CLI list app SwiperTab – Get this Extension for 🦊 Firefox (en-US) GitHub - kouhxp/fftext: Summarize, explain, fact-check, or translate any text, URL, or file. No GPU. No cloud. One command GitHub - sweetpad-dev/sweetpad: Develop Swift/iOS projects using VSCode GitHub - dogmaticdev/IRON: IRON a.k.a. Intermediate Representation Object Notation is a Interpreter/Database that is used to create Programming Languages. GitHub - sjhalani7/vaen: Package your AI coding harness into a portable .agent file, and share it across repos, teams, & the community without ever having to copy-paste instructions, skills, MCP config, or secrets. Show HN: Gandalf the Grader Show HN: Citadeld – replay any CI failure locally from a single file GitHub - tdortman/cuSBF: High-Performance GPU Super Bloom Filter coral-ai/claude-code-token-xray at main · Coral-Bricks-AI/coral-ai GitHub - ulyssestenn/funes: Funes is a Git-based framework for LLM-managed knowledge work: an AI Librarian ingests raw sources, builds an interlinked Markdown knowledge base, and uses it to produce cited reports, analyses, and other outputs. GitHub - ThatXliner/gah: Git Add Hunk, built for agents to use GitHub - harmont-dev/harmont-cli: Command-line client for the Harmont CI platform GitHub - brooksmcmillin/mcp-authflow: OAuth 2.0 Authorization Server framework for MCP servers GitHub - javaid-codes/audit-supply-chain-agents GitHub - amorey/gochan: A small library of common channel architectures for Go, inspired by Rust GitHub - arifozgun/OpenGem: Free, Open-Source AI API Gateway with Gemini, OpenAI & Anthropic Compatibility in 1 file GitHub - Pranesh950/BioPetals: 🌸 Run BIOxAI models at home, BitTorrent-style. Fine-tuning and inference up to 10x faster than offloading GitHub - cnguyen14/bounty-doctor: Diagnose a GitHub bounty issue before you waste hours: detects honeypot scam repos, AI-bot attempt swarms, and stale contests. Show HN: CoreMCP – MCP Server for On-Prem DBs Show HN: KittyHTML – Render HTML/CSS as an inline image in your terminal GitHub - bingud/filemat: Web-based file manager Show HN: TruthLens – Free multi-signal deepfake image detector GitHub - apexlocal-jz/claude-usage-tray: Windows system-tray app showing your Claude Code rate-limit usage at a glance. Zero deps, ~300 lines of PowerShell. Cross-IDE (works regardless of VS Code, Cursor, plain terminal). Release v0.1.2.1 · kouhxp/yapsnap
GitHub - vaddisrinivas/tab-council: Chrome MV3 extension ...
srinivasvadd · 2026-05-29 · via Show HN

Turn the AI tabs you already use into a structured model council.

Tab Council is a Chrome MV3 extension for cross-checking important prompts across ChatGPT, Claude, Gemini, Perplexity, Merlin, and Grok. Create a tab group named tab-council, place your provider tabs inside it, and run one prompt through an independent answer, critique, judge synthesis, and optional Ratify/Veto flow.

No API keys. No backend. Your accounts, your tabs, your local Chrome storage.

Demo

Watch the v0.1.0-alpha demo video

Why It Exists

One model answer is useful. A small council is safer.

There are already browser extensions that send one prompt to many AI tools and show the answers side by side. Tab Council is trying to do something narrower: run the full model-council workflow over the AI tabs you already have open.

It is built for launch reviews, product decisions, architecture checks, research synthesis, policy drafts, and any question where you want independent reasoning plus visible dissent before you trust the answer.

What Makes It Different

  • Uses existing AI tabs: no OpenRouter setup, no provider API keys, no copied model roster.
  • Starts from a Chrome tab group: the tabs you group under tab-council are the council.
  • Runs council rounds, not just broadcast: independent answers, critique, disagreement check, optional final positions, synthesis.
  • Has explicit roles: Member, Judge, Validator, Observer, Exclude.
  • Handles structured output: JSON parsing, one repair pass, and loud failures instead of silent bad transcripts.
  • Adds final review: Ratify/Veto lets non-judge tabs challenge the synthesized verdict.

v0.1.0-alpha

This alpha focuses on extension-based orchestration:

  • Detects the active Chrome tab group named tab-council.
  • Uses each grouped AI tab as an explicit model seat.
  • Reads visible model labels where the provider UI exposes them.
  • Lets every tab take a role: Member, Judge, Validator, Observer, or Exclude.
  • Runs Round 1 independent answers, Round 2 polite critique, optional Round 3, synthesis, and optional Ratify/Veto.
  • Requires structured JSON from every phase.
  • Runs one JSON repair pass when a provider returns almost-JSON instead of valid JSON.
  • Stores run state and history only in chrome.storage.local.
  • Exports a local Markdown transcript.

Supported Providers

  • ChatGPT: chatgpt.com, chat.openai.com
  • Claude: claude.ai
  • Gemini: gemini.google.com
  • Perplexity: perplexity.ai, www.perplexity.ai
  • Merlin: getmerlin.in, www.getmerlin.in, extension.getmerlin.in
  • Grok: grok.com, x.com, *.x.ai
  • Local/dev generic AI pages: localhost and 127.0.0.1

Unknown public web pages are not broadly enabled in the store-safe manifest. Known adapters fail visibly when a provider UI changes.

How It Works

  1. Open the AI provider tabs you want.
  2. Select the model inside each provider tab, such as Claude Sonnet, a ChatGPT model, Gemini, Perplexity, or Grok.
  3. Group those tabs in Chrome.
  4. Name the group tab-council.
  5. Open the Tab Council side panel.
  6. Assign roles and choose a mode.
  7. Enter one prompt.
  8. Review the rounds, final verdict, Ratify/Veto results, and export.

Roles

  • Member participates in debate rounds and can be selected as judge.
  • Judge participates as a member and is preferred for synthesis.
  • Validator skips debate and only ratifies or vetoes the final verdict.
  • Observer appears in the roster but is never automated.
  • Exclude is ignored for the run.

Council Modes

  • Quick: faster pass using current threads.
  • Balanced: fresh threads, critique, judge synthesis.
  • Thorough: forces final statements before synthesis.
  • Rigorous: forces final statements and Ratify/Veto.

Manual controls let you override the mode: force Round 3, choose judge mode, enable Ratify/Veto, include raw export, and select output views.

Structured Output and JSON Repair

Every provider is asked to return strict JSON for each phase. That makes the side panel reliable: confidence scores, disagreements, verdicts, and vetoes can be parsed instead of scraped from prose.

JSON repair means this: if a model returns almost valid JSON, for example markdown fences, trailing comments, or malformed quote escaping, Tab Council sends that same tab a repair prompt and asks for valid JSON only. The repaired response is marked in the UI. If repair fails, the tab fails loud instead of silently corrupting the transcript.

Privacy

Tab Council has no extension-owned server, analytics, or tracking pixels.

  • Prompts are pasted into the provider tabs you selected.
  • Later rounds intentionally share successful model outputs across selected provider tabs.
  • Local run state lives in chrome.storage.local.
  • Exports are generated locally by your browser.
  • Clear Local removes stored extension run data.

Provider websites still receive the content sent into their own tabs. See PRIVACY.md for the exact boundary.

Install the Alpha

Download the v0.1.0-alpha release zip from GitHub:

github.com/vaddisrinivas/tab-council/releases

Then:

  1. Open chrome://extensions.
  2. Enable Developer mode.
  3. Click Load unpacked.
  4. Unzip the release and select the folder containing manifest.json.
  5. Sign into your AI providers in Chrome.
  6. Put those tabs in a group named tab-council.
  7. Click the Tab Council extension icon.

Local Development

npm install
npm run check

Run the deterministic fake provider harness:

Open http://localhost:4173/ in two or more tabs, put them in a tab-council group, grant localhost permission, and run a council. The page returns structured JSON for every phase.

Architecture

  • manifest.json: MV3 permissions, side panel, service worker, provider host allowlist.
  • src/background.js: tab-group discovery, permissions, run state machine, prompt orchestration.
  • src/contentScript.js: provider page automation adapters.
  • src/sidepanel.*: operator UI, roles, run controls, transcript, export.
  • src/prompts.js: structured prompt builders and Markdown export.
  • src/shared.js: provider registry, settings, parsing, helper contracts.

Read more in docs/ARCHITECTURE.md.

Docs