Dev Memory — AI Code Recall

Local-first semantic memory for your codebase and coding agents.

Dev Memory helps developers and AI agents recall implementation details across a repository without sending source code to external services.

What It Offers

• Local semantic indexing of your workspace
• Natural-language code search across files and chunks
• MCP server integration for Copilot and other MCP-compatible agents
• Built-in VS Code welcome UI for indexing and search
• Embedding runtime visibility in the UI (xenova-wasm vs fallback)
• Single-VSIX packaging with bundled WASM runtime dependencies

Why It Is Helpful

• Faster code navigation for large repositories
• Better AI answers grounded in your actual codebase
• Lower context-switching cost when onboarding or switching modules
• Works offline and avoids API key setup
• Keeps sensitive code local by default

Why It Can Be Lucrative

• Teams pay for productivity gains from faster delivery and reduced debugging time
• Local/private-by-default positioning is attractive for enterprise and regulated environments
• MCP compatibility enables integration into modern AI coding workflows
• Foundation for paid tiers: team memory sync, policy controls, analytics, enterprise support

Key Commands

• Dev Memory: Open Welcome Page
• Dev Memory: Index Project
• Dev Memory: Search Project Memory

MCP Tools

• index_codebase
• search_codebase
• summarize_codebase
• save_project_note
• get_embedding_status
• answer_from_codebase

Copilot + MCP Usage

If GitHub Copilot Chat has MCP enabled, select #devmemory-local and ask questions in plain language.

Example prompts:

• Search this codebase for where welcome page command is registered.
• Search this codebase for how indexing works.
• Summarize this repository architecture.
• Answer from codebase: where is search_codebase implemented?

Tech Stack

• TypeScript (monorepo packages)
• VS Code Extension API
• Model Context Protocol (MCP SDK)
• @xenova/transformers (WASM embeddings)
• onnxruntime-web runtime path
• JSON-based local vector index (.dev-memory/index.json)

Branding and Screenshots

Screenshots

Welcome page

Indexing workflow

Semantic search results

Copilot + MCP usage

Packaging Model

Runtime dependencies are bundled under:

• dist/runtime/node_modules/...

This enables one portable VSIX without per-machine native module builds.

Notes

• Indexing is currently explicit/manual.
• Re-index after a VS Code reload before searching.

Privacy

• 100% local processing
• No cloud dependency for core workflow
• No API keys required

License

MIT