atlas-guardrails
v1.0.23
Published
Atlas Guardrails CLI
Downloads
141
Readme
ATLAS GUARDRAILS - Stop the Entropy
Stop LLM agents from turning your codebase into a landfill.

Quick Start • Features • MCP Support • Contributing
The Problem
Coding agents (Claude Code, Cursor, Windsurf) are fast, but they have the memory of a goldfish. They:
- Re-invent the wheel: Creating
utils/date.tswhenlib/time.tsalready exists. - Hallucinate APIs: Guessing method signatures instead of looking them up.
- Drift APIs: Changing public exports without you realizing it until CI explodes.
The Solution: Atlas
Atlas is a local-first guardrail that forces agents to "read the map" before they write code. It indexes your repo, packs relevant context deterministically, and screams at agents when they try to duplicate code or break APIs.
Installation & Integration 🤖
Atlas is built for AI Agents, not humans. To give your agent "eyes" into your repository, choose your client below:
Install Atlas as a native extension:
gemini extensions install https://github.com/marcusgoll/atlas-guardrailsIf installation fails, add it as a manual MCP extension:
gemini mcp add atlas -- command npx -y atlas-guardrails mcpCapabilities added: atlas_index, atlas_pack, atlas_find_duplicates.
Install Atlas as a native skill:
claude extensions install https://github.com/marcusgoll/atlas-guardrailsClaude will automatically utilize SKILL.md and CLAUDE.md context to manage your repo entropy.
Add Atlas as an MCP Server in your IDE settings:
- Open Settings -> Features -> MCP.
- Click + Add Server:
- Name:
Atlas - Type:
command - Command:
npx -y atlas-guardrails mcp
- Name:
Add Atlas to your claude_desktop_config.json:
{
"mcpServers": {
"atlas": {
"command": "npx",
"args": ["-y", "atlas-guardrails", "mcp"]
}
}
}To use Atlas as a standalone terminal tool (or to enable the global atlas command):
npm install -g atlas-guardrailsNote: If you are on a version < 1.0.15, run this one last time to enable auto-updates.
Keep Atlas current with:
atlas updateAgent Workflow
Once installed, your AI agent will follow this deterministic loop:
- Map the Terrain: Agent calls
atlas_indexto build/update the symbol graph. - Gather Context: Agent calls
atlas_packwith your task description. It receives a token-optimized pack of relevant files and their dependency trails. - Prevent Duplication: Before the agent writes a new helper, it calls
atlas_find_duplicatesto see if the code already exists. - Enforce Guardrails: Agent runs
atlas check(or you run it in CI) to ensure no public API drift occurred.
Documentation & Specs
- INTEGRATION.md - Full MCP & API Schema.
- CLAUDE.md - Instruction set for Claude.
- GEMINI.md - Instruction set for Gemini.
- API Documentation - TypeDoc output.
Contributing
We aim for >80% test coverage to keep the guardrails stable.
- Fork & Clone.
npm installnpm test- PR.
License
MIT © Marcus Gollahon
