Log inSubmit server
MCP Architector logo

MCP Architector

Store, search, organize, and repair project architecture facts locally through MCP tools.

InstallRepository
Full README Compare alternatives Embed widget Claim this listing

Install MCP Architector

Pick your client, then add it in one click or paste the config.

STDIONode.js
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "mcp-architector": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-architector"
      ],
      "env": {
        "MCP_PROJECT_ID": "YOUR_VALUE_HERE"
      }
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

Values are written into the snippet above so you can paste it as is. They never leave your browser.

Install not yet sandbox-tested— we haven't run this command through our automated sandbox yet. Not a red flag; we're working through the catalog.

Overview

The mcp-architector MCP server stores project architecture facts, searchable entries, and reusable slices on the local filesystem. It exposes MCP tools for upserting, retrieving, filtering, and deleting entries, plus maintaining slice definitions and repairing the entry index. Data is kept under `~/.mcp-architector`, with project-specific directories and no remote storage described. Reach for it when an agent needs a persistent local reference for architecture decisions, module facts, APIs, or system-design notes.

Use cases

  • Store architecture facts for a project locally
  • Search APIs, domain terms, and system-design notes
  • Group entries into reusable filtered slices
  • Repair a stale entry index or malformed catalog
  • Bulk-import up to 200 architecture facts

Key features

  • Local project-scoped storage
  • Single and bulk entry upserts
  • Filtered entry search with pagination
  • Built-in and custom slices
  • Entry index rebuilding and data repair

Tools12

What MCP Architector lets your agent do. Expand a tool for its parameters.

Self-reported~190 tokens of context

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked 3d ago

Recent health check succeeded.

5/5 checks healthy over the last 46d

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 9d ago via OSV.dev · mcp-architector (npm)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

npm downloads560Package downloads in the last 30 days.
Last commit27d agoMost recent push to the default branch.
Tools exposed12Callable tools this server registers over MCP.

How MCP Architector works

What mcp-architector does

The mcp-architector MCP server gives an AI agent a local store for project architecture information. Its entry model treats each fact as a separate canonical record, such as an API detail, domain term, flow, or other architecture note. Entries can be created, updated, retrieved, deleted, listed, and searched through MCP tools.

The server also provides slices: saved filters that produce focused views over entries. Built-in and custom slices can be listed, while custom slices store filter definitions rather than copied entry content. This lets an agent request a category such as a group of tagged facts without duplicating the underlying data.

How it works

Data is stored beneath ~/.mcp-architector, with a directory for each project. Entry bodies reside in individual files under entries/, while entries/index.json acts as a catalog without duplicating those bodies. Custom slice definitions are stored under slices/.

Use set-entry for one fact or set-entries for a bulk update of up to 200 facts. The bulk operation can attach a moduleName reference to every submitted entry. list-entries returns catalog information, with filters for kind, tags, and text query; search-entries returns compact matches with snippets, slices, module names, and pagination. A matching entry can then be loaded in full with get-entry.

Slices are read views rather than independent data collections. set-slice saves a filter based on kinds and tags, and get-slice applies a slice with optional formatting, search text, limits, and offsets. Results indicate whether more data is available.

Setup and configuration

Install the package globally with npm install -g mcp-architector, or run it through npx. An MCP client configuration can launch it with npx -y mcp-architector. The documented configuration sets MCP_PROJECT_ID to the workspace folder. Claude Desktop uses its platform-specific MCP configuration file; Cursor places the same server block in its Model Context Protocol settings, and Continue uses .continue/config.json.

For development, clone the repository, run npm install, and build with npm run build.

Tools and capabilities

  • Upsert one entry with set-entry, or upsert up to 200 with set-entries.
  • Retrieve a complete record with get-entry and remove one with delete-entry.
  • Browse catalog metadata with list-entries using kind, tag, or query filters.
  • Search compact entry content with search-entries, including module, kind, tag, and pagination filters.
  • Inspect available slices and counts with list-slices, then render a filtered view with get-slice.
  • Create or remove custom kind-and-tag filters with set-slice and delete-slice.
  • Restore the entry catalog with rebuild-entry-index or repair malformed catalog JSON with fix-data.

Limitations and notes

The server is local-first: the documented storage location is on the machine running the server, and no hosted synchronization is described. Direct edits to the storage directory are discouraged because tool-managed timestamps, merge behavior, and index consistency may otherwise be bypassed.

Bulk writes are limited to 200 entries. A slice contains filter information, not item copies, so it does not independently preserve a snapshot of matching entries. When a write response includes a reminder, it can indicate that modules are missing or unlinked. The entry index may need rebuilding if it falls out of sync with the individual entry files.

The mcp-architector MCP server is best suited to agents that need structured, persistent architecture notes rather than a general-purpose database or remote collaboration service.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about MCP Architector

mcp-architector is an MCP server that connects AI clients to local project architecture data stored under ~/.mcp-architector. Its main tools create, update, retrieve, delete, list, and search entries; manage filtered slices; rebuild the entry index; and repair malformed catalog data without requiring a remote service.

Install it with npm or run it directly through npx. Use `npm install -g mcp-architector` for a global installation, or configure an MCP client with command `npx` and arguments `-y`, `mcp-architector`.

Yes, mcp-architector works with Cursor through its Model Context Protocol settings. Add the documented server block under `mcpServers`, launch it with npx or a local command, and set `MCP_PROJECT_ID` to the workspace folder.

It stores project data locally under `~/.mcp-architector`.

Related MCP servers

All in Architecture & Design All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewMCP Architector AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/mcp-architector?style=directory)](https://allmcps.com/mcp/mcp-architector)
HTML Embed
<a href="https://allmcps.com/mcp/mcp-architector"><img src="https://allmcps.com/api/badge/mcp-architector?style=directory" alt="MCP Architector on AllMCPs" /></a>