When AI coding agents work on large projects, they lose architectural context between tasks. Design decisions made early get forgotten, leading to inconsistent implementations and rework. This tool preserves those decisions in persistent context so agents stay aligned across the entire codebase.
Agent Design Decision Memory works as a VS Code extension that watches your agent interactions. Each design decision (e.g., "use dependency injection for database access") gets stored in an indexed, searchable format. Future agent calls pull these decisions automatically into their context window.
Yes. The extension stores decisions in a standardized JSON format that any LLM can ingest. We provide prompt templates for Claude, GPT-4, and local models like Llama. You can also write custom templates for your specific agent.
The VS Code extension is $29/month. Storage for up to 500 decisions and 100 projects is included. Enterprise plans with unlimited storage and team collaboration start at $199/month.
No. The extension stores decisions locally and adds them to the agent context as a single compact block (approximately 2-5% token overhead). The speed gain from agents not re-discovering decisions usually outweighs this small cost.
Yes. You can export your entire decision store as JSON or CSV from the extension settings. We also offer automatic daily backups to S3 or your preferred cloud storage for team accounts.
By default, decisions are stored locally on your machine. You can optionally enable cloud sync to our servers (encrypted) for team collaboration. All data is encrypted at rest and in transit using AES-256.
Yes. Team accounts allow you to share decision stores across developers. Each team member's agent pulls from the shared repository, staying aligned on architecture choices and avoiding divergent implementations.
Yes. All decisions are versioned with timestamp, author, and change summary. You can roll back to any previous decision state. Enterprise plans include full audit logs and decision approval workflows.
After installing the extension, run your agent as usual. Whenever the agent makes a structural choice, highlight the relevant text and click "Save Design Decision" in the context menu. Add a title, category, and optional rationale for future reference.
Yes. We provide a CLI tool that scans your codebase for architecture patterns and suggests decisions for your approval. This typically imports 20-50 decisions in a few minutes.
The extension alerts you when there's a conflict. You can review the contradiction and either update your stored decision or prompt the agent to re-approach the task with the correct architectural context.
Check that the decision category matches what your agent is working on. You can also manually sync by running "Refresh Decision Context" from the command palette. If you're using a local agent, ensure it's in the supported models list.
Yes. Right-click any decision in the tree view and select "Archive" (keeps it in history) or "Delete" (permanent removal). Archived decisions don't appear in agent contexts but remain in your audit log.
Contact our support team at support@agentdesignmemory.dev or check our community Slack channel. We maintain an active support forum and respond to issues within 24 hours.