
feedback-memory
Local memory that recalls your past approvals, edits and rejections of agent drafts
- Pricing
- Open source
- Platforms
- self-hosted, CLI, macOS
- Runs
- Locally
- Version
- 0.2.1
- Listing updated
- 1 day ago
About feedback-memory
feedback-memory is an open-source MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.
Feedback-memory stores each decision you make on an agent's draft, whether approved, edited or rejected, together with a reason and an embedding of the task. Before the next draft, the agent can pull the closest past corrections into its prompt. Plain approvals are stored but not retrieved, since they carry no lesson.
Data lives in a local Postgres with pgvector (PGlite) under ~/.feedback-memory by default, so there is nothing to host. An optional OpenAI key switches matching to text-embedding-3-small embeddings; without it, matching falls back to offline word hashing, where similar means shared words.
The README comes from a one-person education business in Korea that approves agent drafts from a phone. In Claude Code it also ships as a plugin whose hook adds relevant corrections to each prompt even when no tool is called. A browser demo runs the whole loop in a tab.
Install
Run in a terminal
claude mcp add feedback-memory --env OPENAI_API_KEY='<OPENAI_API_KEY>' -- npx -y https://github.com/ssap-pa/self-learning-agent-setup/releases/download/v0.2.1/feedback-memory.mcpbAdd to claude_desktop_config.json
{
"mcpServers": {
"feedback-memory": {
"command": "npx",
"args": [
"-y",
"https://github.com/ssap-pa/self-learning-agent-setup/releases/download/v0.2.1/feedback-memory.mcpb"
],
"env": {
"OPENAI_API_KEY": "<OPENAI_API_KEY>"
}
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"feedback-memory": {
"command": "npx",
"args": [
"-y",
"https://github.com/ssap-pa/self-learning-agent-setup/releases/download/v0.2.1/feedback-memory.mcpb"
],
"env": {
"OPENAI_API_KEY": "<OPENAI_API_KEY>"
}
}
}
}Add to .vscode/mcp.json
{
"servers": {
"feedback-memory": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"https://github.com/ssap-pa/self-learning-agent-setup/releases/download/v0.2.1/feedback-memory.mcpb"
],
"env": {
"OPENAI_API_KEY": "<OPENAI_API_KEY>"
}
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"feedback-memory": {
"command": "npx",
"args": [
"-y",
"https://github.com/ssap-pa/self-learning-agent-setup/releases/download/v0.2.1/feedback-memory.mcpb"
],
"env": {
"OPENAI_API_KEY": "<OPENAI_API_KEY>"
}
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"feedback-memory": {
"command": "npx",
"args": [
"-y",
"https://github.com/ssap-pa/self-learning-agent-setup/releases/download/v0.2.1/feedback-memory.mcpb"
],
"env": {
"OPENAI_API_KEY": "<OPENAI_API_KEY>"
}
}
}
}Configuration
OPENAI_API_KEYOptional. Match by meaning (text-embedding-3-small); empty means offline word hashingFEEDBACK_MEMORY_DIRWhere the database lives (default ~/.feedback-memory)
Tools this server provides (2)
Taken from the project documentation; the running server may differ.
- record_decision
- Record an approve, edit or reject decision on a draft.
- recall_corrections
- Recall the closest past corrections for a new task.
What it does
- Record decisions
- The record_decision tool saves an approve, edit or reject with its reason.
- Recall corrections
- The recall_corrections tool fetches the closest past corrections before drafting.
- Local database
- Uses PGlite with pgvector in a folder on your machine.
- Optional embeddings
- Matches by meaning with an OpenAI key, or by shared words offline.
- Claude Code plugin
- Loads standing rules at session start and injects corrections through a hook.
Good for
- Stop an agent repeating wording you already rejected
- Keep customer-reply drafts consistent with past edits
- Carry standing business rules across sessions
- Try the feedback loop in the browser demo first
Pricing
Open source
Prices change. Check ssap-pa.github.io before you buy.
feedback-memory FAQ
Where is the data stored?
In a local database, by default in ~/.feedback-memory, configurable with FEEDBACK_MEMORY_DIR.
Is an API key required?
No. OPENAI_API_KEY is optional and only enables semantic matching.
Which clients work?
The README names Claude Code, Cursor and Claude Desktop.
Are approvals recalled?
Plain approvals are stored but not retrieved because they carry no lesson.
Is feedback-memory free?
Yes. feedback-memory is open source, so you can use and self-host it for free.
How do I add feedback-memory to Claude Code?
Run this command in your terminal: claude mcp add feedback-memory --env OPENAI_API_KEY='<OPENAI_API_KEY>' -- npx -y https://github.com/ssap-pa/self-learning-agent-setup/releases/download/v0.2.1/feedback-memory.mcpb
What are the best alternatives to feedback-memory?
The closest alternatives to feedback-memory are Codebase Memory, XERJ, Claude Code Ultimate Guide, Nautilus Compass Memory and note1.
Alternatives to feedback-memory

Codebase Memory
MCP serverLocal MCP server that builds a code knowledge graph for AI coding agents
Runs locally- Open source
- 48.7K GitHub stars
- Knowledge and memory

XERJ
MCP serverLocal search engine for AI agents with cited answers over your code and docs
Runs locally- Open source
- 19.2K GitHub stars
- Knowledge and memory

Claude Code Ultimate Guide
MCP serverMCP server for searching the Claude Code Ultimate Guide from any MCP client
Runs locally- Open source
- 6.5K GitHub stars
- AI models and LLM tooling

Nautilus Compass Memory
MCP serverLocal-first long-term memory layer for AI agents, served over MCP
Not checked yet- 5 tools
- Remote
- Open source
- 1.3K GitHub stars
- Knowledge and memory

note1
MCP serverMCP server giving AI assistants search and summaries across your note1 meetings
- 4 tools
- Remote
- Free trial
- 1 GitHub star
- Productivity
MedXpert 图书馆连接器(通用版)
MCP serverLocal MCP server that searches and answers questions over your own Markdown wiki
Runs locally- 4 tools
- Open source
- Knowledge and memory