Gemina
Gemina is a hosted MCP server for working with documents of any kind — invoices, receipts, contracts, forms, statements, delivery notes, or any other PDF or image your team handles. Define your own custom templates to extract exactly the fields you need from any document type, or use designated templates for specific document types. One endpoint, 13 tools in three groups. • FileTag (free tier, 1,500 tags/month, no credit card): turn any PDF or image into structured metadata, six suggested filenames, and a metadata-embedded copy. Upload a file (files_create_upload → tag_file) or point at a public URL (tag_url). • Extraction: OCR for any document, custom templates for your own document types and fields, plus ready-made invoice headers, invoice line items (extract_document, get_extraction_result, list_extractions, get_extraction, get_document, submit_extraction_feedback). • Document Intelligence: search your indexed documents with structured, semantic or hybrid queries (query_documents), run sums, averages and counts grouped by vendor, currency, type or month (aggregate_documents), and re-index documents (index_document). Sign in with your Gemina account — no keys to paste. The free tier covers FileTag; Extraction and Document Intelligence need a paid plan. Supports PDF, PNG, JPEG, GIF, WebP, HEIC/HEIF and AVIF up to 50 MB. You choose the data-center region (EU, US, or Asia) per account, so documents are stored and processed where you decide.
Connection details
13 tool names reported
Source-declared tools. Availability can depend on the connection and permissions; these are not runtime test results.
Claude connector directory ↗ · Observed 2026-11-01 · Completeness not established
aggregate_documentsextract_documentfiles_create_extraction_uploadfiles_create_uploadget_documentget_extractionget_extraction_resultindex_documentlist_extractionsquery_documentssubmit_extraction_feedbacktag_filetag_url
Sources and provenance
- Claude connector directoryClaude tier: community · Observed
Grouping basis: endpoint. Exact endpoints or package sets can be grouped automatically; other relationships require evidence review. This does not verify publisher affiliation, connection health, or equivalent capabilities.