# OpenBMB/UltraRAG

A Low-Code MCP Framework for Building Complex and Innovative RAG Pipelines

Repository: https://github.com/OpenBMB/UltraRAG
Canonical: https://ross.abutalabs.com/products/ultrarag
Homepage: https://ultrarag.github.io/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: embedding, llm, rag, easy, mcp, openai, vllm, gpt, qwen, deepseek, sentence-transformers, demo, flask, ui, huggingface-transformers, multimodal, vlm
Last push: 2026-08-24T05:03:22+00:00

## Health v2 (maintenance only)
Score: 80/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 99, release rhythm 78, longevity 42
- inputs: {"age_days": 594, "days_push": 9, "days_rel": 146, "gap_med": 22, "n_releases_24m": 8}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 5672, forks 440 (observed 2026-08-28T04:09:27.691320+00:00)

## What it is
UltraRAG is a low-code Python framework from OpenBMB for building complex RAG pipelines using MCP servers and YAML-based orchestration with serial, loop, and conditional control flow. It includes a visual UI for knowledge base management, workflow building, and system demos, plus support for multimodal inputs and evaluation.

## Use cases
- build a RAG pipeline over my own documents
- orchestrate retrieval and generation workflows with low-code YAML
- create a visual demo of a RAG system with a knowledge base UI
- run a local DeepResearch-style agent pipeline
- evaluate and iterate on RAG experiments
- build multimodal RAG with vision-language models

## When to choose
- you want to prototype or productionize RAG pipelines without writing lots of glue code
- you need MCP-based tool integration and pipeline control flow (loops, branches)
- you want a bundled UI for demos and knowledge base management
- you work in the OpenBMB/LLM ecosystem with vLLM, Qwen, or DeepSeek models

## When to avoid
- you need a simple drop-in vector database or embedding library rather than a full pipeline framework
- you require a non-Python stack or managed cloud RAG service
- your use case is a single fixed retrieval flow where a few lines of code suffice

## Facets
- artifact type: framework
- maturity: active
- function: rag, agent-framework, mcp, web-framework, llm-inference, data-science
- domain: large-language-models, machine-learning, developer-tools
- platform: python, self-hosted, cross-platform
- tags: low-code, mcp-framework, pipeline-orchestration, yaml-config, multimodal, knowledge-base, deepresearch, retrieval-pipeline, retrieval-augmented-generation, ai-agents, docker, web-server

## Member repositories
- OpenBMB/UltraRAG (main) score 80

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:27.691320+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:54:17.254705+00:00, confidence not recorded.
  - readme: https://github.com/OpenBMB/UltraRAG (fetched 2026-08-28T04:09:27.691320+00:00, sha d2125e19e4a7)
  - homepage: https://ultrarag.github.io/ (fetched 2026-08-29T08:49:25.843589+00:00, sha ddb5fddbb3b1)
- Data as of 2026-08-30T08:39:29.467469+00:00.