Home · Setup · Claude Desktop

MCP Purchase Requisition in Claude Desktop

Purchase requisitions you build once and run per order, with a signed record of each run. You hold a requisition as a named, versioned list of steps, optionally grouped into sections, each required or optional, and run it against an order as many times as you like. A run copies its requisition's steps when it starts and records the version, so editing the requisition afterwards never changes a run already under way. run_check marks steps pass, fail or na with who and when (na counts as answered, never as passed; a required step failed blocks sign-off), run_report gives the whole run with failures in full, and run_sign_off puts a name and a date on it and freezes it. RUN-YYYY-NNNN ids are sequential and never reissued.

The pre-delivery check is usually written in this window, so building the requisition and starting the run for a specific van and PO is one sentence, and the question that matters, can this run be signed off after that failure, is the next one. Nothing leaves the machine: the server makes no network call on any tool.

What you get in Claude Desktop

You sayTool
Make a pre-delivery vehicle check requisition: tyres, lights, load secured.purchase-requisition_create then purchase-requisition_item_add
Start a run of it for van BX21 KLM against PO-0031.run_start
The weight step failed at 3620 against a plated 3500. Can it be signed off?run_check then run_sign_off

The unit suite is 15 tests, 15 passing, 0 failing (node --test, node v22), covering the snapshot rule, the na-is-not-pass rule and the sign-off freeze; adversarial, concurrency, paths and contract suites sit alongside it.

Install it in Claude Desktop

The file is claude_desktop_config.json, and the key inside it is mcpServers.

ScopePath
macOS~/Library/Application Support/Claude/claude_desktop_config.json
Windows%APPDATA%\Claude\claude_desktop_config.json

One click, with no JSON. Download purchase-requisition.mcpb from the latest release and open it. Claude Desktop installs it as an extension and runs it on the Node runtime it ships with, so neither your PATH nor your node version comes into it, and there is no file to edit. Everything below is the manual route, for when you want the entry in the config file itself.

The config block, which needs both strings absolute:

{
  "mcpServers": {
    "purchase-requisition": {
      "command": "node",
      "args": ["/absolute/path/to/mcp-servers/servers/purchase-requisition/dist/index.js"]
    }
  }
}

If node came from nvm, asdf or homebrew, use the absolute path that which node prints rather than the bare word, because a stdio server started by this app inherits a limited PATH:

"command": "/Users/you/.nvm/versions/node/v22.14.0/bin/node"

The node command above wants a built file. From a clone, once:

git clone https://github.com/theluckystrike/mcp-servers.git
cd mcp-servers && npm install
npm run build -w packages/mcp-license -w servers/purchase-requisition

That writes servers/purchase-requisition/dist/index.js, and its absolute path is the one argument the config needs.

The npm packages are not published yet, so the form below returns a 404 today. It is here because it is what the entry becomes the day the publish lands, with nothing else changed:

{
  "mcpServers": {
    "purchase-requisition": {
      "command": "npx",
      "args": ["-y", "@theluckystrike/mcp-purchase-requisition"]
    }
  }
}

Save, then completely quit Claude Desktop and start it again. A window reload is not enough.

Nothing is published to npm yet, so the working install paths are the ones above: the .mcpb bundle from the latest release, a clone and build, or the hosted URL.

No install: the hosted endpoint

The same server runs at https://mcp.zovo.one/mcp/purchase-requisition over MCP streamable HTTP, no install. Mint a free token with curl https://mcp.zovo.one/mcp/token, or use a Pro key. It has no filesystem, so a file comes back as a one-hour download link.

{
  "mcpServers": {
    "purchase-requisition": {
      "url": "https://mcp.zovo.one/mcp/purchase-requisition",
      "headers": { "Authorization": "Bearer <token>" }
    }
  }
}

Claude Desktop notes worth knowing first

Two documented facts decide whether the entry works: every path in claude_desktop_config.json must be absolute, and a stdio server started here inherits only a limited, platform-dependent subset of environment variables. If node came from nvm or homebrew, paste what `which node` prints instead of the bare word. The .mcpb route avoids both: Claude Desktop ships a built-in Node.js environment.

Free: Three requisitions, and unlimited runs of them on every tier, because capping the running of a requisition would cap the only thing a requisition is for. Deleting a requisition frees its slot and every run stays readable. Pro is $19 once, verified offline.

Questions

Where is claude_desktop_config.json?

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json. Windows: %APPDATA%\Claude\claude_desktop_config.json. There is no published Linux path. Settings, Developer, Edit Config opens the right copy and creates it if it does not exist.

I added it and MCP Purchase Requisition does not appear.

In order: did you fully quit and restart, not just close the window? Is every path absolute? And is the command resolvable, given the limited environment a stdio server inherits here? If node came from nvm, paste what which node prints.

Can I install it without editing JSON?

Yes. An .mcpb bundle, the format renamed from .dxt, opens with Claude and shows an installation dialog. One you built yourself goes in through Settings, Extensions, Advanced settings, Install Extension.

Related

MCP Purchase Requisition in detail · The same server in Claude.ai and Claude Desktop connectors and Claude Code · Every server in Claude Desktop · Guides · Claude Desktop docs · Buy Pro