Limits

Generated from the exported limits object of @sezzlee/pdf-mcp 0.1.2.

Every limit is fixed at build time; none is configurable. A call that would cross one either answers with truncated: true and a way to continue, or fails with resource_limit.

Files and extraction

Limit	Value
Largest document the server opens	32 MB
Most pages a document may have; a longer one is refused before it is extracted	2,000
Time one extraction may take before the call fails with `resource_limit`	20 s
Documents that may be extracted at once before a new read fails with `resource_limit`	2
Extracted documents kept in memory between calls	2
Directory entries `list_documents` walks before it reports `scanTruncated`	5,000
Documents `list_documents` returns when `maxResults` is omitted	50
Largest `maxResults` for `list_documents`	200
`list_documents` calls that may run at once	4

Responses

Limit	Value
Largest serialized answer; a longer page is clamped and continues through `nextCursor`	512 KB
Pages `read_pages` returns when `maxPages` is omitted	10
Largest `maxPages` for `read_pages`, and the most page numbers `pages` may name	50

Search

Limit	Value
Matches `find_in_document` returns when `maxResults` is omitted	50
Largest `maxResults` for `find_in_document`	200
Longest `context` returned around one match	240
Longest `query`, in characters	256

OCR

Limit	Value
Resolution pages are rendered at when the binding sets no `dpi`	200
Pages sent to the provider in one call; a binding's `maxPagesPerCall` can lower it, never raise it	10
Time one OCR run may take; a binding's `timeoutMs` can lower it, never raise it	120 s
OCR runs that may be in progress at once before a new one fails with `resource_limit`	1
Transcribed pages kept in memory, keyed by document content and page number	256