MCP tool reference

The ProcDuck MCP server offers 242 tools, covering nearly everything you do in the app: searching tenders, pricing items, quotations, supplier RFQs, sales and purchase orders, invoices and the team. This reference explains each tool: what it does, its parameters, what it returns and who may use it.

How the tools work

Each tool matches an action you take in the app: search tenders, price an item, create a quotation, record a supplier quote. A tool's name starts with the area and ends with the action, like tenders_list, quotations_create or rfq_responses_add. You rarely type a tool name yourself: you ask Claude for what you want in your own words, and it picks the right tool and calls it as you. If you are building your own MCP client, these names and parameters are what you send.

Every call runs as the member who connected, inside their organization, and goes through the same REST API (v1) the app and your own scripts use, so no tool goes around your permissions. The REST endpoints behind the tools are in the API reference.

Who may use a tool

A call needs three things at once, and the line under each tool's description names them:

• The scope: a permission written on the API key or connection, such as tenders.read or quotations.write. Reading and writing are separate scopes, and * grants everything. A connection made by signing in from Claude carries *, bounded by your role.
• The role: your current role in the organization. It is read on every call, so a change to your role changes the connection at once. Some roles use a tool without seeing prices, and the notes under the tool say so.
• The plan: every tool needs the Pro or Enterprise plan. Starter includes no API or MCP access, and every request from it answers 403 plan_upgrade_required with the feature api_access. On Pro each user may hold one API key or MCP connection, and a second is refused with 403 plan_limit_reached and api_keys_per_user. Enterprise has no limit.

A refusal says why: a 403 whose text names the missing scope, or starts with insufficient_role, plan_upgrade_required or plan_limit_reached. The detail of each role is in the API reference.

Reading an entry

• The line under the description: first what the tool does to your data (read-only, changes data, or deletes or overwrites data), then the scope, the roles and the plans.
• The parameters table: each parameter's name, type, whether it is required, and what it means. Nested fields are written with a dot, so items[].quantity is the quantity field of each element of the items list. The types are string, integer, number, boolean and object, and [] after a type means a list of it.
• Example arguments: what a client actually sends, for those building their own.
• What it returns, and notes on the rules and limits worth knowing before you call.

Conventions

• Listings take page (from 1) and page_size, and each tool's description gives its maximum. An empty result means the search found nothing.
• Updates are partial: an update tool changes only what you send, and whatever you leave out stays as it is. null clears a field that can be empty. A list (such as a manual tender's items) is replaced only when you send it, and a tool's description says when it replaces the whole list.
• IDs are numbers (a tender ID, an item ID, a quotation ID and so on) and come from the list and detail tools. Amounts are in SAR unless the description says otherwise.
• Dates: last_offer_date is a tender's submission deadline and submission_date is when it was published. A date is YYYY-MM-DD and a timestamp is ISO 8601.
• Uploads travel as base64 inside the JSON, in content_base64, or as a reference to a file the platform already holds (staged_ref) where the tool says so. The server refuses a request larger than about 4.5 MB.
• Limits: 100 requests a minute per key, and every call counts against your organization's daily allowance: 1,000 on Pro, unlimited on Enterprise.

Files and background jobs

A tool that generates a file (a PDF, an Excel workbook, a ZIP or XML) returns filename, content_type, size_bytes and the content as base64 in base64. A file over 4 MB is not embedded in the reply: the tool answers with an error that names the REST endpoint to download it from with your key. When the assistant on WhatsApp is the caller, files are sent into the conversation as documents after the reply.

Two files take more than one request to build: the NUPCO attachments file and the quotations file bundle. They are built in the background; the tool that starts one waits up to wait_seconds and then returns the file or the job, and the tools in Bid files and background jobs follow it.

Uploading a file to a record, such as a product photo or attachment or the organization's logo, does not go through the tool: the start tool returns a temporary upload link you send the file to with a PUT, and the finalize tool checks the file and files it. The details are in Photos, attachments and spreadsheet imports.

The pages of this reference

All tools, A to Z