# MiroFlow

🏆 Top-1 on 5+ benchmarks | Web UI | Supports MiroThinker, Claude, Kimi, OpenAI

Repository: https://github.com/MiroMindAI/MiroFlow
Canonical: https://ross.abutalabs.com/products/miroflow
Homepage: https://miromindai.github.io/MiroFlow/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: agent-framework, agents, claude, deep-research, futurex, gaia, gpt-5, hle, research-agent, xbench, browsecomp
Last push: 2026-07-06T14:50:43+00:00
Link (homepage): https://miromindai.github.io/MiroFlow/

## Health v2 (maintenance only)
Score: 59/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 91, release rhythm 35, longevity 28
- inputs: {"age_days": 393, "days_push": 58, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3103, forks 325 (observed 2026-08-28T04:07:43.638794+00:00)

## What it is
MiroFlow is a performance-first open-source agent framework for building, orchestrating, and benchmarking LLM-powered agents, with MiroThinker as its flagship deep research agent model. It supports pluggable LLM providers, multi-agent graph orchestration, a skill system, and reproducible evaluation across benchmarks like GAIA, BrowseComp, HLE, and FutureX.

## Use cases
- build a deep research agent that browses the web and answers complex questions
- benchmark and compare different LLMs on agent tasks like GAIA and BrowseComp
- orchestrate multi-agent workflows with sequential agent graphs
- run reproducible agent evaluations from config files
- create custom agent skills without changing framework code
- self-host a web app for research report generation

## When to choose
- you need a research/search agent with state-of-the-art benchmark results
- you want fair, reproducible head-to-head LLM agent comparisons
- you want to plug any LLM provider into an agent framework via config
- you need multi-agent orchestration with rollback and retry robustness

## When to avoid
- you need a simple chatbot with no research or tool-use requirements
- you want a lightweight agent loop with minimal dependencies
- your workload is unrelated to agentic research or benchmarking

## Facets
- artifact type: framework
- maturity: active
- function: agent-framework, llm-inference, rag, web-scraping, benchmarking, search-engine
- domain: artificial-intelligence, large-language-models, developer-tools
- platform: python, cross-platform
- tags: deep-research, research-agent, llm-agents, multi-agent-orchestration, agent-benchmarks, tool-use, web-ui, ai-agents, retrieval-augmented-generation, web-server, docker

## Member repositories
- MiroMindAI/MiroFlow (main) score 59
- MiroMindAI/MiroThinker (backend) score 59

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:43.638794+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 2026-08-29T17:27:34.051776+00:00, confidence not recorded.
  - readme: https://github.com/MiroMindAI/MiroFlow (fetched 2026-08-28T04:07:43.638794+00:00, sha af4152d6ebe4)
  - homepage: https://miromindai.github.io/MiroFlow/ (fetched 2026-08-29T08:27:31.671739+00:00, sha 9b35b4d50ea5)
- Data as of 2026-08-30T08:39:29.467469+00:00.