Skip to main content
Glama

@aistoragedepot/mcp

A Model Context Protocol server that exposes your AIStorageDepot library to any MCP-aware AI client — Claude Desktop, Cursor, Cline, Windsurf, and others.

Your stored content shows up where you actually work:

  • Prompts → the items you mark with / in the app become MCP prompts (slash-commands) — any type, with each prompt's [field] surfaced as a prompt argument you fill in.

  • Resources → every item (rules, docs, skills, MCP configs, prompts) is readable at aisd://item/<id>, so an agent can pull "my coding rules" or "my project guidelines" into context on demand.

  • Toolssearch_library(query) and get_item(id) let an agent find and fetch anything in your library programmatically.

Try it with no account (sample mode)

AISD_TOKEN is optional. Run the server with no token and it serves a free, read-only sample of the public curated library — a good way to see AIStorageDepot inside your editor before signing up. Add a token later to switch to your own & team libraries. Add the server with an empty env (or omit AISD_TOKEN):

{
  "mcpServers": {
    "aisd": { "command": "npx", "args": ["-y", "@aistoragedepot/mcp"] }
  }
}

Related MCP server: @prompthub/mcp

Setup

👉 Easiest path: step-by-step instructions for each app (config-file locations and all) are at https://www.aistoragedepot.com/connect. The manual version:

  1. In AIStorageDepot, go to Settings → API tokens and create a token (free for individuals). Copy it.

  2. Add the server to your MCP client's config. Example (Claude Desktop / claude_desktop_config.json, Cursor ~/.cursor/mcp.json, etc.):

{
  "mcpServers": {
    "aisd": {
      "command": "npx",
      "args": ["-y", "@aistoragedepot/mcp"],
      "env": {
        "AISD_TOKEN": "aisd_xxxxxxxxxxxxxxxxxxxxxxxxxxxx"
      }
    }
  }
}
  1. Restart the client. The items you marked / appear as slash-commands; everything is available as resources; and the search_library / get_item tools are ready.

Configuration

Env var

Required

Default

Notes

AISD_TOKEN

no

Token from Settings → API tokens. No token = anonymous sample mode (the free public curated library, read-only). With a token you get your own & team libraries. Has full access to your account — keep it secret.

AISD_BASE_URL

no

https://www.aistoragedepot.com

Point at a self-hosted / dev instance if needed.

AISD_WORKSPACES

no

Retired. Your slash menu is now the items you mark / in the app (strict opt-in); this variable is ignored. Search and resources still cover every workspace.

Slash-commands (pull)

Not every client turns MCP prompts into / slash-commands (some only use the tools). To get your library as native slash-commands, sync it to command files — for every AI tool you use, not just Claude:

npx -y @aistoragedepot/mcp pull --token=aisd_your_token --to=all

Target

Writes to

Commands appear as

claude (default)

~/.claude/commands/aisd/

/aisd:<name>

cursor

~/.cursor/commands/

/aisd-<name>

vscode

your VS Code profile's prompts/ folder (Copilot Chat)

/aisd-<name>

windsurf

~/.codeium/windsurf/global_workflows/ (Cascade)

/aisd-<name>

codex

~/.codex/prompts/

/prompts:aisd-<name>

Tools that aren't installed are skipped automatically. Each tool gets its own dialect (frontmatter, argument hints, $ARGUMENTS where supported). Skills run as-is. Prompts take input: their [fields] show as an argument hint, and whatever you type after the command fills them —

/aisd:cold-outreach Jane at Globex, about our payments API

— with the AI asking for any value you left out.

Option

Default

--to=…

claude

Which tools to write commands for — comma list, or all.

--token=…

AISD_TOKEN env

Your API token.

--types=…

skill,prompt

Which item types become commands (e.g. skill for skills only).

--workspaces="A,B"

your personal library

Which workspaces to pull (names, or all).

--project

off

Write into the current repo instead: .claude/commands/, .cursor/commands/, .github/prompts/, .windsurf/workflows/.

--dir=<path>

per-target

Write somewhere else entirely (single --to only).

Re-runs are safe: each target only clears its own files (claude: its namespaced folder; the others: only files with the aisd- prefix — your own commands are never touched).

Notes

  • Read-only in v0.1 (it surfaces and fetches your content; it doesn't write back).

  • The library snapshot is cached for ~30s, so a burst of calls hits the API once.

  • Revoke a token any time in Settings → API tokens; the server loses access immediately.

Local development

npm install
AISD_TOKEN=<token> AISD_BASE_URL=http://localhost:3100 npm run test:client

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections. Dates show when Glama detected each change.

  1. 2 tool updatesv0.6.7
    • Removedget_item
    • Removedsearch_library
  2. 2 tool updatesv0.1.1
    • Changedget_item1 field changed
      • changedInput schema / properties / id / description
        Previous value: -"The item id, or an aisd://item/<id> URI."New value: +"An item id, or a full aisd://item/<id> URI — the aisd://item/ prefix is stripped for you, so search_library output can be pasted unchanged. Exact match only, one id per call: no titles, slugs, partial ids, wildcards, or comma-separated lists, so fetch N items with N calls. Use only an id returned by search_library, a resource listing, or supplied by the user — ids cannot be derived from titles, and an id outside the libraries you can read resolves as not found."
    • Addedsearch_library
  3. 1 tool updatev0.1.0
    • First observedget_item

TDQS

A4.8/5.0
Disambiguation5/5

The two tools have completely distinct purposes: search_library discovers items by keyword (returns metadata only) and get_item retrieves full content by ID. There is no overlap or ambiguity in their functionalities.

Naming Consistency5/5

Both tool names follow a consistent verb_noun pattern (search_library, get_item) using snake_case, making it clear what each tool does at a glance.

Tool Count4/5

With only 2 tools, the server is minimal but well-scoped for a read-only library search and retrieval workflow. While one might expect additional tools for browsing or management, the count fits the stated purpose.

Completeness3/5

The server covers the core search and retrieve workflow, but lacks tools for browsing libraries without a keyword or listing all items. This limits discoverability and leaves some common operations unsupported.

Maintenance

ActivitySlowing
ResponsivenessSyncing

Related MCP Connectors

Related MCP Servers

  • A
    license
    A
    quality
    A
    maintenance
    Turns team chat (Slack, Discord, MS Teams, Mattermost) into a living knowledge base - a typed knowledge graph plus auto-generated wiki - exposed to AI agents through a 28-tool MCP server with semantic search, expert finding, and decision tracing.
    28
    443
    Apache 2.0
  • A
    license
    Not graded
    quality
    B
    maintenance
    Self-hosted MCP server for storing and serving structured team knowledge, enabling AI sessions to load relevant team context without re-prompting.
    Apache 2.0
  • A
    license
    A
    quality
    D
    maintenance
    Delivers on-demand knowledge from a built-in library of 728 agents, skills, patterns, and examples across 26 domains to any MCP-compatible AI tool.
    6
    14
    1
    MIT

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/AIStorageDepot/aistoragedepot-mcp'

If you have feedback or need assistance with the MCP directory API, please join our Discord server