Skip to main content
Glama
cognitivemyriad

Verified Repo Memory

README.md
![Verified Repo Memory Banner](assets/banner.png)

# Verified Repo Memory MCP v0.1.2

[![CI](https://github.com/cognitivemyriad/mcp-verified-repo-memory/actions/workflows/ci.yml/badge.svg)](https://github.com/cognitivemyriad/mcp-verified-repo-memory/actions/workflows/ci.yml)
[![TypeScript](https://img.shields.io/badge/TypeScript-5.7.0-blue.svg)](https://www.typescriptlang.org/)
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)

Stale-proof repository memory with citations + just-in-time verification + TTL (repo-scoped).

An MCP server providing "safe memory" for AI coding agents. Memories are scoped per repository, backed by code citations, and verified just-in-time so an agent never receives stale information when the underlying code has changed.

## Quickstart

Run via `npx`:

```bash
npx -y @cognitivemyriad/vrm-local --repo /path/to/repo
```

*(Alternatively, run from source: `npm ci && npm run build && node build/index.js --repo /path/to/repo`)*

## Tools

* **`vrm_store`**: Store a new memory with file citations.
* **`vrm_search`**: Search for candidate memories by keywords.
* **`vrm_retrieve`**: JIT-verify candidates and return only valid memories. (Main tool for agents)
* **`vrm_list`**: List memories by status (valid, stale, missing).
* **`vrm_forget`**: Manually delete a memory.

### Example I/O

**Store:**
Input:
```json
{
  "subject": "API version sync",
  "fact": "When changing API version, update client/server/docs together.",
  "citations": [{ "path": "src/api.ts", "startLine": 10, "endLine": 15 }]
}
```
Output:
```json
{
  "stored": true,
  "memoryId": "uuid-...",
  "expiresAt": "2026-03-21T00:00:00Z"
}
```

**Retrieve:**
Input:
```json
{ "query": "API version" }
```
Output:
```json
{
  "query": "API version",
  "valid": [ ... ],
  "stats": { "verified": 1, "validCount": 1 }
}
```

## How it works

```mermaid
graph TD
    A[Agent] -->|Store Fact + Citation| B(Verified Repo Memory)
    B --> C{Save to Disk}
    C -->|Hash Code Snippet| D[(memories.json)]
    
    A -->|Retrieve Fact| B
    B --> E{JIT Verification}
    E -->|Check File Hash| F{Unchanged or Relocated?}
    F -->|Yes| G[Return VALID Memory]
    F -->|No| H[Return STALE/MISSING]
```

1. **Citations:** Every fact is linked to a file path and a line range. The exact code snippet is hashed and saved.
2. **JIT Verification:** Before returning a memory to the agent in `vrm_retrieve`, the server checks the physical file. If the snippet has moved, it relocates the citation. If it has been changed or deleted, the memory is marked STALE/MISSING and omitted from the results.
3. **TTL (Time-To-Live):** Memories expire automatically (default 28 days) unless they are successfully retrieved and utilized, which extends their life.

## Data location

Data is strictly repo-scoped and saved in:
`<repoRoot>/.verified-repo-memory`

This includes `memories.json` and a fingerprint/metadata file to prevent accidental cross-repo pollution. Add this directory to your `.gitignore`.

## Security

* **No Network Transmissions:** This is a `stdio` local-only server without HTTP calls.
* **Path Security:** Disallows any path traversal (`../`) out of the repository root, as well as accessing `.git/` or `.verified-repo-memory/`.
* **No Stdout Pollution:** Strict logging only to `stderr`.
* **Secret Scan:** Built-in heuristic secret scanning to reject memories that look like API keys/private keys (can be disabled via `--no-secret-scan`).

## Usage with Claude

### Claude Desktop
To add this server to the Claude Desktop app, edit your configuration file:
- On macOS: `~/Library/Application Support/Claude/claude_desktop_config.json`
- On Windows: `%APPDATA%\Claude\claude_desktop_config.json`

Add the following configuration:
```json
{
  "mcpServers": {
    "verified-repo-memory": {
      "command": "npx",
      "args": [
        "-y",
        "@cognitivemyriad/vrm-local",
        "--repo",
        "/absolute/path/to/your/repo"
      ]
    }
  }
}
```

### Claude Code
To add this server to Claude Code using `stdio` transport:

```bash
claude mcp add mcp-verified-repo-memory --transport stdio -- npx -y @cognitivemyriad/vrm-local
```

*Note for Windows users:* You may need to prepend `cmd /c` to the command:
```bash
claude mcp add mcp-verified-repo-memory --transport stdio -- cmd /c npx -y @cognitivemyriad/vrm-local
```

---

## Publishing Guide

This section explains how to publish the package to **NPM** and register it with the **Anthropic MCP Registry** so that it becomes publicly available.

### Step 1: Publish to NPM

NPM (Node Package Manager) is the package distribution platform. Publishing here allows anyone to install your tool with a single command.

#### Prerequisites

- An [NPM account](https://www.npmjs.com/signup) (free)
- Two-factor authentication (2FA) enabled on your NPM account
- Node.js installed on your machine

#### Procedure

**1. Log in to NPM from the terminal:**

```bash
npm login
```

A browser window will open. Sign in with your NPM account. When prompted, enter your 2FA code from your authenticator app.

**2. Publish the package:**

```bash
npm publish --access public
```

This command does the following:
- Compiles TypeScript → JavaScript (`npm run build`)
- Creates a `.tgz` archive of the compiled files
- Uploads the archive to `https://registry.npmjs.org/`

**3. Verify the publication:**

Visit `https://www.npmjs.com/package/@cognitivemyriad/vrm-local` in your browser. Your package page should appear.

> **Note:** If you need to re-publish, you must increment the version number in `package.json` and `server.json` first (`npm version patch`). NPM does not allow overwriting existing versions.

---

### Step 2: Register with Anthropic MCP Registry

The [MCP Registry](https://registry.modelcontextprotocol.io) is Anthropic's official directory of MCP servers. Registering here allows Claude Desktop, Claude Code, and other MCP clients to discover and install your server.

> **Important:** The NPM package **must be published first** (Step 1). The MCP Registry validates that the NPM package exists before accepting the registration.

#### Prerequisites

- A [GitHub account](https://github.com) (used for authentication only)
- The `mcp-publisher` CLI tool

#### Installing mcp-publisher

```bash
# macOS (Homebrew)
brew install nicholasgriffintn/tap/mcp-publisher

# Or via npx (no install required)
npx @anthropic-ai/mcp-publisher
```

#### Procedure

**1. Log in to the MCP Registry via GitHub:**

```bash
mcp-publisher login github
```

A browser window will open. Authorize the application with your GitHub account.

**2. Publish to the MCP Registry:**

```bash
mcp-publisher publish
```

This command reads `server.json` in the current directory and registers the server with the MCP Registry. The registry will:
- Validate the `server.json` schema
- Check that the NPM package exists and is accessible
- Register the server metadata (name, description, version, environment variables)

**3. Verify the registration:**

Visit `https://registry.modelcontextprotocol.io` and search for your server name.

> **Note:** Once a version is published to the MCP Registry, it is **immutable** and cannot be changed. To publish updates, increment the version in both `package.json` and `server.json`, publish to NPM first, then run `mcp-publisher publish` again.

---

### Version Management

When releasing a new version, always update the version number in **all three locations**:

```bash
# 1. Bump version in package.json
npm version patch  # 0.1.2 → 0.1.3

# 2. Update server.json (both top-level and packages[].version)
# Edit server.json manually to match the new version

# 3. Publish
npm publish --access public
mcp-publisher publish
```

| File | Field | Must Match |
|------|-------|-----------|
| `package.json` | `version` | ✅ |
| `server.json` | `version` (top-level) | ✅ |
| `server.json` | `packages[0].version` | ✅ |

TDQS

B3.2/5.0

Scored across 5 tools

Disambiguation5/5

Each tool targets a distinct action: store (add), search (find candidates quickly), retrieve (verify and return valid memories), list (enumerate by status), forget (delete). The boundaries are clear, especially with the verification distinction between search and retrieve.

Naming Consistency5/5

All tool names follow the consistent pattern 'vrm_' prefix plus an imperative verb (store, search, retrieve, list, forget). This is a uniform, predictable convention that makes the tool set easy to navigate.

Tool Count5/5

With exactly 5 tools, the server covers the core memory lifecycle (create, read, list, delete) without bloat. Each tool earns its place, and the scope is well-matched to the stated purpose.

Completeness4/5

The set covers the essential operations for a memory store: add, search, retrieve, list, and forget. However, there is no update operation, and retrieval is tied to verification rather than direct ID lookup, which could be a minor gap for some workflows.

Maintenance

ActivityInactive
ResponsivenessNo issues