ssap-pa/self-learning-agent-setup

`feedback-memory` turns your approve, edit and reject decisions into memory an agent checks before it drafts: `record_decision` saves the edit and your reason, `recall_corrections` returns the closest past corrections and standing rules for the next task. Five tools (`recall_corrections`, `record_decision`, `add_rule`, `list_memory`, `forget`); local Postgres + pgvector via PGlite in `~/.feedback-memory`, offline hashing embedder by default or OpenAI embeddings when `OPENAI_API_KEY` is set. Install: `claude mcp add feedback-memory -- npx -y github:ssap-pa/self-learning-agent-setup`. Transport: `stdio`; no auth; clients: Claude Code, Cursor, Claude Desktop, and compatible MCP clients. MIT. In-browser demo: https://ssap-pa.github.io/self-learning-agent-setup/.

Project

Browse this MCP server in the TensorBlock MCP Index.