Pluribus

Security & Safety 💻 JavaScript ⚖️ MIT 🟡 Quiet lately
13 stars

Pluribus is a CLI that generates privacy-safe context receipts for AI coding agents, proving which instruction files, skills, MCP schemas, or memory results actually crossed an agent boundary without logging raw content. It also syncs a single pluribus.md source of truth into native context files for multiple AI tools. It is for teams that need evidence of context fidelity and drift across AI-assisted development workflows.

✨ Key features

  • Generates privacy-safe context receipts for agent boundaries
  • Syncs pluribus.md to CLAUDE.md, .cursorrules, AGENTS.md, and more
  • Audits context files with --fidelity-report for drift and fallbacks
  • Provides demos for MCP traffic, memory, and tool surface evidence
  • Supports imports for shared team context via # @import
  • Includes browser demo and public evidence ledger

🎯 Use cases

  • Prove which context file an agent actually loaded before a task
  • Maintain a single source of truth for project instructions across tools
  • Audit context drift and generic fallback usage in AI coding agents
  • Generate receipts for MCP tool usage without exposing raw schemas
  • Track memory or RAG answer authority and freshness

📦 Installation

🧰 Requirements: Node.js and npm are required to run the CLI commands.

npm install -g pluribus-context

Or run with npx: npx --yes pluribus-context@latest audit --json --fidelity-report

🚀 Usage

Initialize pluribus.md and sync to native files
pluribus init
pluribus sync
Audit context files with fidelity report
npx --yes pluribus-context@latest audit --json --fidelity-report
Run a demo receipt (e.g., MCP traffic)
npm exec --yes --package github:caioribeiroclw-pixel/pluribus -- pluribus demo mcp-traffic-receipt --json

⚠️ Good to know

Pluribus is not a persistent memory layer, retrieval system, agent orchestrator, enterprise ContextOps platform, or agent-merging framework; it only provides evidence for context boundaries.

❓ FAQ

What does Pluribus do exactly?

It generates privacy-safe receipts showing which context files, skills, MCP schemas, or memory results crossed an agent boundary, and it can sync a single pluribus.md into native files for multiple AI tools.

Which AI tools does Pluribus support?

It supports Claude Code, Cursor, GitHub Copilot, OpenClaw, Windsurf, Continue, Zed, Bob, Cline, Roo Code, Amazon Q Developer, JetBrains Junie, Warp, and Gemini CLI.

Does Pluribus log raw prompts or source code?

No, it avoids logging raw prompts, source code, tool output, paths, transcripts, secrets, or customer data; receipts are privacy-safe.

How do I run an audit?

Use npx --yes pluribus-context@latest audit --json --fidelity-report to see native discovery surfaces, generic fallbacks, load evidence, and semantic differences.

📊 Repository

🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.