MCP

MCP tools reference

Tools#

Tool	What it does
`stirling_describe_operation`	Returns the parameters and JSON schema for one operation.
`stirling_pages`	Page operations: merge, split, rotate, rearrange, crop, delete, overlay, add blank pages.
`stirling_convert`	Conversions to and from PDF (Word, images, HTML and more).
`stirling_misc`	Compress, OCR, stamp, watermark, metadata, flatten, repair and similar.
`stirling_security`	Passwords, redaction, sanitising, certificate signing and signature validation.
`stirling_upload`	Stores a file on the server and returns a `fileId`, so several operations can reuse it.
`stirling_download`	Returns a stored file by `fileId`, as base64, up to the inline limit.
`stirling_ai`	Runs AI capabilities. Needs AI to be set up.

Operation IDs are the short endpoint names, such as compress-pdf. Each tool lists the operations it has, and an admin can hide some. Filters and the Pipeline API aren't available over MCP.

Calling a tool#

Call stirling_describe_operation with an operation first to get its parameters. Then call the category tool:

Argument	Required	Value
`operation`	Yes	The operation ID, for example `compress-pdf`.
`parameters`	No	An object of the operation's parameters.
`file`	No	The input document as base64. The usual way to send a file.
`fileName`	No	The original file name with its extension.
`fileId`	No	A file stored with `stirling_upload`, or an earlier result, instead of `file`.

json

{"name": "stirling_misc", "arguments": {"operation": "compress-pdf", "fileName": "report.pdf", "file": "<base64>", "parameters": {"optimizeLevel": 5}}}

Results:

A JSON result comes back as text.
A file result comes back inline as base64 (stirling://file/{fileId}) with its fileId, up to the server's inline limit (10 MB by default on self-hosted servers).
A larger result returns only a fileId, which you can pass to the next operation. stirling_download has the same limit, so it can't return that file either. On a self-hosted server an admin can raise the limit (MCP server settings).

AI capabilities#

stirling_ai takes an operation (a capability ID), a parameters object and, for capabilities that read a document, a fileId. The capabilities are pdf-question-answer, pdf-edit-plan, pdf-comment-generate, agent-draft, agent-revise, agent-next-action, math-audit-examine and math-audit-deliberate. Use stirling_describe_operation to get each one's schema.

AI is on for Stirling Cloud. On a self-hosted server an admin must set it up (AI on your server).

Edit this page on GitHub