Make your agentic AI work like a 100x more effective developer. Maintain a local, permanent memory of database schemas, code conventions, completed milestones, and future tasks—100% offline in your repository.
Starting a new chat wipes all context. Stale decisions lead to database errors and stack mismatch hallucinations. anchor.md provides a permanent local memory layer.
Status: Repetitive & Leaky
Status: Standardized & Versioned
Watch how anchor.md changes every stage of your AI workflow — from setup to daily use.
Without anchor.md, your AI resets to zero on every new chat.
Scenarios
Click any tab to explore each scenario → or run npx create-anchor-md now
Get 100x efficiency immediately. Copy these pre-built workflows directly into Cursor, Claude, ChatGPT, or Gemini.
Use this prompt when bootstrapping a brand new app. It directs the AI to read your rough ideas in @anchor.md/main.md and ask 3-5 critical clarifying questions about the architecture before code is written.
Read my rough idea and tech preferences inside @anchor.md/main.md. idea : [here is given else ask from user]. Before you generate the rest of the workspace or write any code, analyze the project's type, scale, and requirements. Ask me 3-5 critical, tailored questions about the tech stack, auth library, database schema, or target constraints so we can align first.
No databases, no SaaS fees. Just raw Markdown files organized with absolute precision.
Works with Claude, ChatGPT, Gemini, DeepSeek, Copilot — any AI with local file access.
All files live in your repo. No cloud sync, no SaaS subscriptions, no data leaks.
Plain markdown files. Read them, edit them, version-control them using standard Git.
Strict synchronization rules prevent race conditions in multi-agent environments.
6 pre-built expert role prompts (Backend, Frontend, DevOps, QA, Designer, Security).
Switch AI models mid-task without repeating or losing a single context instruction.
Select any file in the structure to view its purpose and typical Markdown layout structure.
The project's entry point. Fill this once with your tech stack, preferences, and architecture goals. The AI reads this first and uses it to bootstrap all other files.
# ⚓ @main.md
## 🚀 PROJECT DETAILS
- **Project Name:** School ERP
- **Type:** Web App (SaaS)
- **Primary Tech Stack:** Next.js, Postgres, Prisma, Tailwind
## 📂 PROJECT STATUS
- [ ] New Project
- [x] Existing Project (Migrations pending)
## 🎯 KEY PREFERENCES
- Strict TypeScript & clean layout
- Modular, decoupled server actionsRun commands directly in your project root to align your local state or bundle files for web AI assistants.
$npx create-anchor-md🚀 Initializing anchor.md context system... Added anchor.md/ to .gitignore Created .gitignore and added anchor.md/ ✅ anchor.md system initialized successfully! Created context directory at: ./anchor.md
Set it up in minutes. The system integrates naturally with whatever IDE chat tool you prefer.
Run the single-line command in your repository root. This copies the lightweight template folder.
Fill in your stack, preferences, and project guidelines inside the main entrypoint file.
Tag @main.md or the role prompts in Cursor, Claude, or Copilot. The AI instantly consumes your repository rules.
The AI updates TODO.md and CONTEXT_MEMORY.md as it writes code, maintaining perfect state.
If your development environment or AI assistant has the ability to read, reference, or attach local files, it fully supports anchor.md out of the box. No custom extensions, configuration, or plugins needed.
Keep your AI aligned on requirements, packages, schemas, and guidelines across every branch and commit.
Get notified about new role prompts, template updates, CLI releases, and core AI development workflows.