# Global Agent Memory

Provides local Markdown-backed shared memory for Claude Code and other agents, with a human review queue in a web dashboard.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI · ★ 14
- Author: ozankasikci
- License: MIT · policy=link-only · description only; the file stays with the author
- Source: https://github.com/ozankasikci/global-agent-memory
- Updated: 2026-09-01
- HTML: https://affpapa.org/en/nevibecode/mcp/global-agent-memory

## What it does

Global Agent Memory connects coding assistants like Claude Code and Codex to an Obsidian-compatible Markdown vault. Agents do not write directly to permanent records. Instead, they submit context as candidate notes, leaving approval, edits, or rejection to the developer through a local browser dashboard.

Search runs on SQLite FTS5 for local keyword matching, with optional semantic embeddings via Ollama and `sqlite-vec`. Background daemons are optional because CLI operations execute in-process when the service is idle. When running the dashboard, the process binds strictly to `127.0.0.1` with a bearer token.

Running the package via `uvx` requires Python 3.12 or newer on your machine. Sensitive context can also be marked as protected, stopping agents from reading it without owner confirmation.

## Requirements
- Needs: Python and uv (commands run via uvx)

## Install

### Claude Code (native platform)

**Platform CLI**
```bash
claude mcp add global-agent-memory -- uvx global-memory-mcp
```
uvx runs the command named after the package; if the README names another one, use uvx --from global-memory-mcp <command>

### Codex

**Platform CLI**
```bash
codex mcp add global-agent-memory -- uvx global-memory-mcp
```
uvx runs the command named after the package; if the README names another one, use uvx --from global-memory-mcp <command>

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "global-agent-memory": {
            "command": "uvx",
            "args": [
                "global-memory-mcp"
            ]
        }
    }
}
```
uvx runs the command named after the package; if the README names another one, use uvx --from global-memory-mcp <command>

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "global-agent-memory": {
            "type": "stdio",
            "command": "uvx",
            "args": [
                "global-memory-mcp"
            ]
        }
    }
}
```
uvx runs the command named after the package; if the README names another one, use uvx --from global-memory-mcp <command>

### Gemini CLI

**Platform CLI**
```bash
gemini mcp add global-agent-memory uvx global-memory-mcp
```
uvx runs the command named after the package; if the README names another one, use uvx --from global-memory-mcp <command>

## When to use it
- share architectural context across different coding agents
- review proposed agent notes before storing them permanently
- retain project conventions in plain Markdown files
- search past bug resolutions with local keyword and vector indexes

## Example prompts
- Look up our database migration rules in project memory.
- Propose a memory candidate for our standard error response format.
- Search the vault for how we resolved the Docker socket permission issue.

## Similar
- [sqlite-graph-memory](https://affpapa.org/en/nevibecode/mcp/sqlite-graph-memory.md) — MCP server · ★ 12
- [context-mem](https://affpapa.org/en/nevibecode/mcp/context-mem.md) — MCP server · ★ 19
- [Engram: AI memory you can edit in Obsidian](https://affpapa.org/en/nevibecode/mcp/engram-ai-memory-you-can-edit-in-obsidian.md) — MCP server · ★ 12
- [State Memory MCP](https://affpapa.org/en/nevibecode/mcp/state-memory-mcp.md) — MCP server · ★ 39
- [neverinfamous-memory-journal-mcp](https://affpapa.org/en/nevibecode/mcp/neverinfamous-memory-journal-mcp.md) — MCP server · ★ 20
- [Lians Agent Memory](https://affpapa.org/en/nevibecode/mcp/lians-agent-memory.md) — MCP server · ★ 16

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md