What it does
Pluribus is a command-line tool (pluribus-context on npm, pluribus in the terminal) for the instruction files that AI coding tools read. You write project context, conventions and constraints once in a versioned pluribus.md, optionally importing shared team Markdown, and sync generates the native files for each tool, such as CLAUDE.md, .cursorrules, Copilot instructions, AGENTS.md and rules for Windsurf, Continue, Zed, Bob, Cline, Roo Code, Amazon Q Developer, Junie, Warp and Gemini CLI. You can check before anything is written. The audit command is read-only and lists existing context files or reports generated files that are missing or have drifted, validate checks the source, and sync --dry-run previews the output. The init command writes only pluribus.md, and generated files carry a header marking them as generated. Normal runs make no network requests; remote imports are fetched only when you pass --update-imports. The project also centres on evidence. It produces receipts that record which instruction files, skills, MCP tool schemas, memory results or compaction steps crossed an agent boundary, without logging raw prompts, source code, tool output or secrets. The page says it is not a memory layer, retrieval system or agent orchestrator. It is released under the MIT license.
Capabilities
- One versioned pluribus.md as the source of truth for project context
- Generates native instruction files for Claude Code, Cursor, Copilot, Windsurf, Zed and more
- Read-only audit and sync --dry-run commands to preview before writing
- Reports generated files that are missing or have drifted
- Privacy-safe context receipts that avoid raw prompts, source code and secrets
- Optional imports of shared team Markdown
When to use it
- You use several AI coding tools and keep near-duplicate rule files for each
- You want to catch drift between instruction files in pull requests or before commits
- You want evidence of which context an agent actually loaded
