All activity
AI coding agents have amnesia. They solve the same bugs repeatedly, forgetting hard-won knowledge the moment the session ends.
PeKG is a cross-project knowledge graph that agents query via MCP. No prompts to write — the agent captures decisions, patterns, and gotchas as you work, compiles them into searchable wiki articles, and surfaces relevant blockers before you hit them again.
BYOLLM: Your agent does the synthesis. PeKG owns the graph. You keep control of your data and tokens.
PeKGAI memory for coding agent: your knowledge, always available
Dana Mercerstarted a discussion
PeKG - Stop re-explaining your codebase to AI every session
The problem: AI coding agents are goldfish. Every session resets. You waste hours re-explaining context your past self already figured out. The solution: PeKG - a personal knowledge graph for developers. Capture decisions, patterns, and fixes as you work. PeKG compiles them into searchable articles and surfaces them before you even ask. Key features: Cross-project intelligence (knowledge...
Dana Mercerleft a comment
PeKG fixes this by giving your coding agent a persistent memory across all your projects. It works via MCP (Model Context Protocol), so it plugs into Claude Code, Cursor, Windsurf, and others with zero config. The twist: we don't run your LLM. Your agent does the heavy lifting (BYOLLM). We just own the graph, search, and compilation pipeline. Would love your feedback — what knowledge do you...
PeKGAI memory for coding agent: your knowledge, always available