Morphik Core

Morphik Core is the open-source multimodal retrieval engine built by Morphik, Inc., offered to developers as a standalone platform at dev.morphik.ai. It powers Morphik's commercial AI back-office product for skilled nursing and senior living operators, but is also available independently for developers building AI applications that need to work with complex, visually rich documents. The project has accumulated over 3,600 GitHub stars since its creation in November 2024, reflecting active community interest.

What It Is

Morphik Core is an AI-native toolset for ingesting, searching, transforming, and managing unstructured and multimodal documents. It addresses a core problem in production RAG (Retrieval-Augmented Generation) pipelines: traditional approaches that work in proof-of-concepts often fail when documents contain charts, diagrams, tables, or mixed visual and text content. Rather than requiring developers to stitch together separate systems for OCR, embeddings, vector databases, and retrieval, Morphik Core provides an end-to-end platform with a single API surface. It is written primarily in Python and licensed under the Business Source License 1.1, with each version automatically re-licensing to Apache 2.0 four years after its first release.

Core Capabilities

Morphik Core's feature set is built around making complex document retrieval production-ready:

• Multimodal Search via ColPali: Uses the ColPali technique to build search that understands the visual content of documents — images, PDFs, videos, and more — through a single endpoint, preserving spatial relationships that text-only pipelines lose.
• Fast and Scalable Metadata Extraction: Extracts metadata from documents including bounding boxes, labeling, and classification.
• MCP Support: Built-in support for the Model Context Protocol, allowing integration with any MCP client in a single click.
• User and Folder Scoping: Multi-user and folder-based access controls for isolating data by project or user group within a unified database.
• Flexible Model Registry: Register and use hundreds of AI models across an application, mixing and matching by task through a simple configuration file.
• App Provisioning: Create isolated apps with their own Morphik URIs for per-customer or per-project data separation.

Developer Access and Setup

Developers can access Morphik Core in two ways. The recommended path is signing up for the hosted cloud service at dev.morphik.ai, which includes a free tier and compute-usage-based pricing for higher ingestion volumes. Self-hosting is also supported via direct installation or Docker, with configuration managed through a morphik.toml file. The Python SDK makes ingestion and querying straightforward — ingesting a file requires just two lines of code. A REST API, TypeScript SDK, and a web-based Morphik Console are also available. Cookbooks in the documentation cover agent workflows, complex metadata filtering, and generating completions with retrieved chunks.

Licensing and Source Availability

Morphik Core is source-available under the Business Source License 1.1, not a traditional OSI-approved open-source license. Personal and indie use is free. Commercial production use is free for deployments generating less than US $2,000/month in gross revenue; above that threshold, a commercial license is required. Each code version automatically converts to Apache 2.0 four years after its first public release — the Change Date for the initial version (0.1.0) is June 18, 2029. The repository is publicly available and accepts community contributions via GitHub issues and pull requests.

Relationship to Morphik's Commercial Product

The GitHub README makes the product lineage explicit: Morphik Core is the retrieval engine behind Morphik's commercial AI back-office platform targeting skilled nursing and senior living operators. The commercial product automates accounts payable, billing, collections, and payroll workflows, and is described on morphik.ai as live across 150+ facilities. Morphik Core is offered separately to developers as a standalone retrieval platform, decoupled from the vertical-specific commercial application.