Releases · CytrexSGR/persistent-agent-memory

persistent-agent-memory v0.1.0

MCP server that gives AI agents persistent semantic memory across sessions.

11 MCP tools — remember, recall, forget, remember_decision, remember_rule, search_knowledge, get_context, get_session_summary, get_bootstrap_context, index_directory, search_indexed
Pluggable storage — SQLite + sqlite-vec (default, zero-config) or PostgreSQL + pgvector
Pluggable embeddings — any OpenAI-compatible API (Ollama, vLLM, OpenAI) or local sentence-transformers
Directory indexer — semantic search over Markdown files / Obsidian vaults with heading-based chunking and re-index detection
Session continuity — handoff + bootstrap skills for structured agent transitions
Multi-agent ready — source agent tracking, typed categories (decision, rule, handoff, project)

pip install persistent-agent-memory
ollama pull nomic-embed-text
pam serve --stdio

Add to Claude Code .mcp.json:

{"persistent-memory": {"command": "pam", "args": ["serve", "--stdio"]}}