Migrate from Merlin to Twenty CRM

PDF, PPT, and Word files uploaded in Merlin Document Chat migrate as file attachments in Twenty. We preserve file binary content, original filename, and upload timestamp. Each attachment is linked to a parent Company or People record that we create during entity extraction. Free-tier Merlin documents (capped at 10 pages) are flagged separately so customers understand if truncation occurred during the original upload. Pro-tier unlimited-length uploads migrate in full.

Merlin Document Chat does not store contacts natively. We perform document-level parsing on all uploaded PDFs, presentations, and Word files to extract person names, email addresses, job titles, and phone numbers. Each extracted entity becomes a Twenty People record. Multiple mentions of the same person across documents deduplicate by email or fuzzy name match and consolidate into a single People record with a source_documents field listing each source file.

Company names and domains extracted from Merlin uploaded documents map to Twenty Companies. We extract company domain from email addresses found in documents, match against known company name patterns, and create a Companies record per unique organization. The Twenty Companies domain field is populated where extractable from document content.

Merlin chat sessions migrate as Twenty Note records. Each session becomes a Note with the conversation transcript as body text, the session start timestamp as ActivityDate, and a link to the most relevant Company or People record via ContentDocumentLink. Sessions exceeding Twenty's Note body limit split into sequential Notes with part numbering. Model attribution metadata (which AI model responded per turn) is not exported from Merlin and is flagged as permanently unavailable.

Merlin Document Chat Q&A pairs (the question asked against a specific uploaded file and the AI-generated answer) migrate as Twenty Notes with a tag indicating the source file. We preserve the question text, answer text, and source document reference. This allows sales reps to review the document analysis that informed past decisions without re-uploading the file to Twenty.

Merlin account metadata (email, plan tier, usage quota) from account settings maps to Twenty Members. The Merlin Owner assignment for conversations is not a first-class concept, so owner attribution on Notes relies on session export timestamps matched against the account holder's email. Teams with multiple Merlin users require separate account exports and individual Member provisioning in Twenty before migration.

Merlin image generation outputs migrate as file attachments in Twenty. We preserve the PNG or image binary, creation timestamp, and prompt text as a description field. Image attachments link to the relevant Company or People record if the prompt context allows association; otherwise they attach to a general Workspace or Opportunity record.

Merlin has no custom property schema. Any structured data that appears consistently across uploaded documents (e.g., project codes, contract values, product SKUs) is identified during entity extraction and mapped to a Twenty Custom Object created in Settings before migration. We document the field names, data types, and source documents for each custom entity so the customer's admin can configure the schema in Twenty's Data Model section.

Merlin has no pipeline, deal, or opportunity objects. We do not migrate pipeline stages because none exist in the source. If the customer has deal context embedded in Merlin documents, we extract deal-related entities (deal name, value, stage name, expected close date) as Custom Object records in Twenty rather than native Opportunity records.

Merlin does not expose a tagging or labeling system accessible via export. We cannot migrate structured tag metadata. Any labels visible in Merlin's UI are documented as a manual reference for the customer to recreate in Twenty's tagging or topic features post-migration.