{
  "markdown": "<p align=\"center\">\n  <img src=\"public/voyage/voyage-mast.jpg\" alt=\"Argus — Odysseus bound to the mast before the sirens' song\" width=\"820\">\n</p>\n<p align=\"center\"><em>\nOdysseus bound himself before the Sirens sang.<br>\nArgus binds a decision to its premises and return condition before fluency — or hindsight — can rewrite them.\n</em></p>\n\n<h1 align=\"center\">Argus</h1>\n\n<p align=\"center\"><strong>Keeping Judgment Human.</strong></p>\n<p align=\"center\">Turn a decision into movement. Let reality shape the next call.</p>\n\n<p align=\"center\">\n  <a href=\"https://www.npmjs.com/package/argus-decision-mcp\"><img src=\"https://img.shields.io/npm/v/argus-decision-mcp?color=A8842F&label=npm%20%C2%B7%20argus-decision-mcp\" alt=\"npm version\"></a>\n  <img src=\"https://img.shields.io/badge/Claude%20Code%20plugin-argus-667572\" alt=\"Claude Code plugin\">\n  <img src=\"https://img.shields.io/badge/license-open--core-6E8261\" alt=\"license: open-core\">\n  <img src=\"https://img.shields.io/badge/local--first-no%20account%20needed-242321\" alt=\"local-first\">\n</p>\n\n<p align=\"center\">\n  <a href=\"https://argus.voyage\"><strong>Web app</strong></a> ·\n  <a href=\"#-try-it-in-30-seconds\">Install</a> ·\n  <a href=\"#what-argus-does\">What it does</a> ·\n  <a href=\"#the-judgment-record\">The record</a> ·\n  <a href=\"#license\">License</a> ·\n  <a href=\"./README.ko.md\">한국어</a>\n</p>\n\nArgus turns a stuck decision into one observable next move, keeps the words and\nassumptions you actually adopted, and brings the record back when reality can\nanswer. The durable output is not another AI answer: it is a closed,\nappend-only loop that can improve the next call without scoring the person.\n\n---\n\n## Why \"Argus\"?\n\n<img align=\"right\" width=\"210\" src=\"public/images/brand/argus-v2/argus-returning.jpg\" alt=\"Argus — the hound who knew Odysseus beneath the disguise\">\n\nThe mast above and the name **Argus** carry the two halves of the product.\nOdysseus bound himself before the Sirens could overtake his judgment. His hound,\nArgus, recognized him twenty years later beneath a beggar's rags. One holds a\nchoice steady against persuasion; the other preserves memory, recognition, and\nfidelity beyond time and appearances.\n\nThat's the whole product. Work with an AI long enough and its smooth voice\nquietly takes over your memory of *why* you decided. Argus is the dog at the\ndoor. It **remembers what you believed** when you decided, **notices the hidden\npremise** under the fluent answer, **keeps watch** on the signals you chose, and\n**returns first** when reality has an answer.\n\nIt never decides for you, and it never grades the person. Reality can answer a\nclaim; only you can answer what happened to your commitment or standard.\n\n---\n\n## ⚡ Try it in 30 seconds\n\nThree doors, one loop. Start with whichever fits.\n\n| | Best for | Get started |\n|---|---|---|\n| 🌐 **Web app** | Anyone. Nothing to install, no signup. | Open **[argus.voyage](https://argus.voyage)** |\n| 🧩 **MCP server** | Any AI assistant that speaks MCP — Claude Code, Claude Desktop, Codex, Cursor… | `claude mcp add argus -- npx -y argus-decision-mcp`<br><sub>other hosts → [argus-mcp/README.md](./argus-mcp/README.md)</sub> |\n| 🔌 **Agent Plugin** | The same loop inside agent hosts; Claude Code also gets the full review suite. | Agent Plugins: install `argus-plugin-v2/`<br>Claude Code: `/plugin marketplace add commet/Argus` then `/plugin install argus@argus` |\n\n<sub>New and not sure? The **web app** is the durable Decision Desk and needs zero setup. Want Argus inside existing AI work? Use the **MCP server** or portable **Agent Plugin**. Full setup + tool list in **[argus-mcp/README.md](./argus-mcp/README.md)** and **[argus-plugin-v2/README.md](./argus-plugin-v2/README.md)**.</sub>\n\n<sub>**What you need:** the web app needs only a browser. The MCP server and the plugin need **Node.js 18+** on your `PATH` — check with `node --version`, and get it from [nodejs.org](https://nodejs.org) if that prints nothing (20 LTS is the tested version). No API key, no account, no config file: records are local files from the first use.</sub>\n\n---\n\n## What Argus does\n\nTwo things — and the second is the one most tools skip.\n\n### ① While you decide — it keeps *your* judgment separate from the AI's\n\nWork with an AI long enough and its assumptions quietly blend into yours. Argus\npulls them apart and keeps them honest:\n\n- **Remember exactly.** Your words, your premises, your confirmed call — and *whose*\n  each one is. Every premise stays tagged as yours or the AI's, so something the\n  model surfaced never quietly becomes your belief.\n- **Notice beneath the surface.** It surfaces the hidden premise and the real\n  question under the fluent answer — without choosing a side for you.\n\n### ② After you decide — it keeps watch and returns when the question is answerable\n\n- **Keep watch honestly.** It tracks only the signals *you* chose, and re-checks\n  the load-bearing premises while your decisions are still open. If several\n  decisions depend on the same premise, one material change brings them together\n  in a single re-check — showing the connection instead of stacking duplicate\n  alerts.\n- **Return first.** It comes back on the date you set — or earlier, when a watched\n  premise materially moves.\n- **Keep the chronology, not a verdict.** Argus shows the original first. You\n  append what reality showed, what happened to the commitment, and whether the\n  original question still makes sense. No score is stored.\n\n---\n\n## One loop\n\nEverywhere Argus runs, the loop is the same:\n\n1. **Record & separate** — keep the first utterance, the confirmed wording, and\n   whether an AI suggestion was adopted.\n2. **Make it answerable** — choose what kind of question it is and a useful event\n   or fallback date. A moment kept only for the record needs no return.\n3. **Wait quietly** — no nagging; if the named event happens or the date arrives,\n   Argus says so once.\n4. **Return original-first** — you answer reality, commitment, and question\n   validity separately. The model never invents the answer.\n5. **Keep the chronology** — later wording and answers append. Earlier wording\n   stays intact, and no person-level score or win rate is stored.\n\n<details>\n<summary>What that looks like at each door</summary>\n\n- **Web app** — write the decision in one line (or upload a strategy doc); an AI\n  crew surfaces the hidden premises, marks what only a human can call, and alerts\n  you when a premise changes.\n- **MCP server** — talk about a decision naturally in any chat; Argus preserves\n  the user-authorized record and pulls it back on its event or fallback date.\n- **Claude Code plugin** — five commands over your real code and PRs:\n  `/argus:review` (deep pressure-test, explicit opt-in) · `/argus:check` ·\n  `/argus:history` · `/argus:settings` · `/argus:help`.\n\n</details>\n\n---\n\n## The Judgment Record\n\nThe artifact across time. Notice the split: **what the AI proposed** stays apart\nfrom **the wording you authorized**. A later answer appends; it does not rewrite\nthe earlier statement.\n\n```\n┌─ ARGUS · JUDGMENT RECORD ─────────────────────────────────┐\n  Confirmed 2026-04-02       Returned 2026-06-30\n\n  FIRST WORDS\n    \"Can we cut over without a maintenance window users notice?\"\n  AI-SURFACED ASSUMPTION\n    The index rebuild fits inside the replication lag budget.\n  MY CONFIRMED CALL\n    \"Proceed only if measured downtime stays under 5 minutes.\"\n  RETURN CONDITION\n    After the production cutover · fallback 2026-06-30\n\n  REALITY         Cutover took 3 minutes, no customer reports.\n  STANDARD NOW    Still the same.\n  QUESTION        Still valid.\n  ─────────────────────────────────────────────────────────\n  NO PERSON-LEVEL SCORE OR WIN RATE IS STORED.\n└────────────────────── argus · statement → return → answer ┘\n```\n\n---\n\n## How it differs from answer tools\n\n| | Answer tools | **Argus** |\n|---|---|---|\n| Output | a more plausible answer | an **append-only judgment record** |\n| Whose reasoning it keeps | the model's answer | **yours**, with AI proposals and adoption lineage kept separate |\n| Who evaluates | the model (scores, confidence) | **nobody scores the person**; reality and the user's later answer remain separate facts |\n| After the chat ends | the reasoning evaporates | Argus **comes back** — named events plus a fallback date |\n| What accumulates | an implicit profile of you | **a chronology you own**: original wording, revisions, conditions, observations, and answers |\n\n---\n\n## An honest promise about your data\n\n- **Local first.** The ledger lives in the current project's `.argus` directory (MCP/plugin) and your browser\n  (web). Uninstalling never deletes your records.\n- **No human score — structurally.** There is no scoring tool to call, and new\n  foundation records reject score-shaped fields at the storage boundary.\n- **No silent profiling.** Derived patterns about you are excluded from prompts by\n  default; anything that would use them needs your explicit, scoped, revocable\n  grant — and leaves a visible trace.\n\n---\n\n## Setup in detail\n\n### 🌐 Web — use it right away\n\nNo install, straight in the browser: **[argus.voyage](https://argus.voyage)**.\nNo account required.\n\n### 🧩 MCP server\n\nAdd Argus to any AI assistant that supports MCP (Claude Code, Claude Desktop, Codex,\nCursor, …). The fastest path, in Claude Code:\n\n```bash\nclaude mcp add argus -- npx -y argus-decision-mcp          # this project\nclaude mcp add -s user argus -- npx -y argus-decision-mcp  # every project\n```\n\nZero config works — your ledger lives in the current project's `.argus`. For the full config (Claude\nDesktop, Windows, per-project ledgers, optional account sync) and the six tools,\nsee **[argus-mcp/README.md](./argus-mcp/README.md)**.\n\n### 🔌 Claude Code plugin\n\nOne install wires everything — the skills, the MCP server, and quiet reminders:\n\n```text\n/plugin marketplace add commet/Argus\n/plugin install argus@argus\n```\n\nRestart Claude Code, then from any repo:\n\n```text\n/argus:review \"the question you need to decide\"\n```\n\n**Commands (5):** `/argus:review` (deep pressure-test, explicit opt-in) ·\n`/argus:check` (settle what is due · seal · premises) · `/argus:history`\n(decision log · version tree · scan past chats) · `/argus:settings` ·\n`/argus:help`. These are the complete public commands. More →\n**[argus-plugin-v2/README.md](./argus-plugin-v2/README.md)**.\n\n<details>\n<summary>Local development &amp; optional API key</summary>\n\n```bash\ngit clone https://github.com/commet/Argus.git\ncd Argus\nnpm install\nnpm run dev            # runs at http://localhost:3000\n```\n\nSee [CONTRIBUTING.md](./.github/CONTRIBUTING.md) for the repo layout, the CI checks, and\nthe conventions reviewers look for.\n\n**API key (optional, web).** By default the web app works through a server proxy.\nTo use it without rate limits, pick \"Direct Key\" mode on the settings page and\nenter your Anthropic API key — it is stored only in your browser's localStorage\nand is never sent to our server.\n\n</details>\n\n---\n\n## Repository layout\n\n```\nsrc/               # the web app (Next.js — argus.voyage)\nargus-mcp/         # the MCP server (npm: argus-decision-mcp, MIT)\nargus-plugin-v2/   # the Claude Code plugin (marketplace: argus, MIT)\ndocs/ARGUS-BLUEPRINT.md   # the build canon (what gets built, in what order)\n```\n\nDetails → [CONTRIBUTING.md](./.github/CONTRIBUTING.md).\n\n---\n\n## License\n\nArgus is **open-core**: the parts meant to spread are open source, and the product\nitself is source-available. This repo is **licensed by part** (full explanation →\n[LICENSING.md](./LICENSING.md)):\n\n| Part | License | Commercial use |\n|---|---|---|\n| `argus-plugin-v2/`, `argus-mcp/` (plugin · MCP) | **MIT** (open source) | ✅ Freely |\n| Everything else — the **web app core** (`src/`, etc.) | **PolyForm Noncommercial 1.0.0** | ❌ Needs a separate commercial license |\n\nIn other words, **the plugin and the MCP server are open source (MIT)** and can be\nused freely, including commercially. **The web app source is source-available** —\nyou can read it, learn from it, and run it personally, but **commercial use is not\npermitted**. For a commercial license, open an [issue](https://github.com/commet/Argus/issues).\n\nTrademarks: the \"Argus\" name, logo, and argus.voyage are trademarks and are not\ncovered by the licenses above.\n\n---\n\n<p align=\"center\">\n  <em>Argus translates a way of thinking hidden inside the strategy-planning craft<br>into a form anyone can use — and keeps the judgment where it belongs: with you.</em>\n</p>\n",
  "bytes": 12464,
  "sha": "c820d6d2d7bbd91534f12eef0f4448431e9f45f04d42164893eef680d440f991",
  "repo_slug": "commet/argus",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/mcp_io_github_commet_argus_decision_mcp_8128120c/readme"
}