riponcm/projectmem

Overview

A local-first memory layer for AI coding agents that captures issues, attempts, decisions, and cross-project library gotchas. It runs as a native MCP server verified across Claude Desktop, Cursor, Antigravity, and Codex, keeping all data 100% local with no cloud or telemetry.

Best for

Best for
Developers who want persistent, private memory for AI coding agents across sessions.

Use cases

• Persisting AI agent context across coding sessions to avoid repeating past discoveries
• Recording cross-project library pitfalls so agents learn from related work
• Storing decision logs for complex debugging or refactoring tasks

Notes

A local-first memory layer for AI coding agents that captures issues, attempts, decisions, and cross-project library gotchas. It runs as a native MCP server verified across Claude Desktop, Cursor, Antigravity, and Codex, keeping all data 100% local with no cloud or telemetry.

19 stars on GitHub. Last updated 2026-06-18. Licensed MIT.

Use cases

• Persisting AI agent context across coding sessions to avoid repeating past discoveries
• Recording cross-project library pitfalls so agents learn from related work
• Storing decision logs for complex debugging or refactoring tasks

Pros

• No cloud dependency or telemetry — fully offline and private
• Works with multiple agent frontends via standard MCP protocol
• Low overhead: Python-based and easy to integrate into existing workflows

Cons

• Small community (19 stars) — limited long-term support or roadmap
• Requires manual setup and configuration of the MCP server
• No built-in sharing or syncing between multiple machines

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.