Alternatives to Docling
The closest alternatives to Docling are Marker, Unstructured and Firecrawl: they are also libraries and share Retrieval-Augmented Generation, Document Ingestion, Document Processing with it. 14 direct alternatives were found by overlap of capabilities and integrations (top score 0.75), plus 8 adjacent solutions of a different product type. As of 2026-11-21. DER
Subject: Docling · Library · IBM (LF AI & Data) · licence MIT · 71k GitHub stars
Direct alternatives (14)
Same product type. Similarity is the share of capabilities and integrations the two have in common out of all either has (0–1).
| # | Alternative | Similarity | Shared with Docling | Licence | Hosting | Stars | Last push | Differs in | Compare |
|---|---|---|---|---|---|---|---|---|---|
| 1 | MarkerDatalab | 0.75 | Apache-2.0 | Unknown | 41k | 2026-11-20 | licence Apache-2.0, fewer stars | Head-to-head | |
| 2 | UnstructuredUnstructured | 0.63 | Apache-2.0 | Unknown | 16k | 2026-11-19 | licence Apache-2.0, fewer stars | Head-to-head | |
| 3 | Firecrawlfirecrawl | 0.60 | AGPL-3.0 | Unknown | 206k | 2026-11-21 | licence AGPL-3.0, more stars | Head-to-head | |
| 4 | RAGFlowinfiniflow | 0.60 | Apache-2.0 | Unknown | 94k | 2026-11-20 | licence Apache-2.0, more stars | Head-to-head | |
| 5 | Crawl4AIunclecode | 0.60 | Apache-2.0 | Unknown | 88k | 2026-11-20 | licence Apache-2.0, more stars | Head-to-head | |
| 6 | LlamaParseLlamaIndex | 0.60 | MIT | Unknown | 68 | 2026-11-20 | fewer stars | Head-to-head | |
| 7 | GraphRAGmicrosoft | 0.60 | MIT | Unknown | 37k | 2026-11-19 | fewer stars | Head-to-head | |
| 8 | LightRAGHKU Data Intelligence Lab | 0.60 | MIT | Unknown | 41k | 2026-11-21 | fewer stars | Head-to-head | |
| 9 | MarkItDownmicrosoft | 0.60 | MIT | Unknown | 197k | 2026-11-07 | more stars | Head-to-head | |
| 10 | Jina Readerjina-ai | 0.60 | Apache-2.0 | Unknown | 12k | 2026-05-22 | licence Apache-2.0, fewer stars | Head-to-head | |
| 11 | TavilyTavily | 0.60 | Unknown | Unknown | Unknown | Unknown | licence unknown | Head-to-head | |
| 12 | ExaExa | 0.60 | Unknown | Unknown | Unknown | Unknown | licence unknown | Head-to-head | |
| 13 | ReductoReducto | 0.60 | Unknown | Unknown | Unknown | Unknown | licence unknown | Head-to-head | |
| 14 | VectaraVectara | 0.60 | Unknown | Unknown | Unknown | Unknown | licence unknown | Head-to-head |
The competitor graph keeps at most the 10 closest products per subject (similarity ≥ 0.34). "Differs in" compares licence, hosting and stars against Docling.
Adjacent solutions (8)
Overlapping capabilities, but a different product form: the same job done as a skill, plugin, library or service instead.
| Product | Type | Similarity | Shared | Stars |
|---|---|---|---|---|
| DocSlicer | MCP server | 1.00 | 36 | |
| unfold-mcp | MCP server | 0.71 | 0 | |
| Carbone MCP | MCP server | 0.67 | 6 | |
| office-oxide-mcp | MCP server | 0.67 | 165 | |
| Kami | Agent skill | 0.67 | 13k | |
| office-transform | Agent skill | 0.57 | Unknown | |
| add-anydoc | Agent skill | 0.57 | Unknown | |
| carbone-skill | Plugin | 0.57 | 6 |
Head-to-head comparisons
How alternatives are determined
For every profiled product, the pipeline takes its specific capabilities (keyword- or researcher-assigned taxonomy nodes, not inherited parents) and the integrations named in its metadata, and computes the Jaccard overlap with every other profiled product: shared items divided by all items either has. Pairs scoring at least 0.34 are kept, the 10 best per product. Same product type is a direct alternative; a different type is adjacent.
Limits: capabilities come mostly from keyword rules on publisher descriptions, so a product with a short or unusual description can be missed or mis-paired. Similarity says the descriptions cover the same ground, not that the products are interchangeable, equally good or tested together. Stars are GitHub popularity for the linked repository. AgentSidekick has not run any of these products.
See the methodology, or compare Docling with the top three in the compare tool.
Page data last updated .