{"repo":"frontier-ai-next/flowMAS","free":true,"listed":false,"github":"https://github.com/frontier-ai-next/flowMAS","clone":"git clone https://github.com/frontier-ai-next/flowMAS.git","description":"Graph-native studio, runtime integration, and self-hosted observability for mutable multi-agent LLM workflows.","language":"TypeScript","stars":20,"topics":["agentic-ai","fastapi","gmas","graph","llm","multi-agent-systems","observability","react","workflow-automation"],"license":"MIT","category":"ai-agents","readme_excerpt":"flowMAS Graph-native web studio, demo API, and self-hosted observability for mutable multi-agent workflows powered by gMAS . Live demo · Documentation · Core gMAS SDK · Evaluation · MIT license - Workflow studio — visual graph editor: agents, edges, conditions, execution-order preview - Run monitoring — live timeline, tools, tokens, events, topology changes - Configure — LLM providers, tools, runner config, routing policy, topology hooks - Observe — run history, metrics, diagnostics, error output The gMAS SDK core lives in a separate repository ( frontier-ai-next/gMAS ) and is vendored here as a git submodule at vendor/gmas . This repository contains the studio, API, and observability integration; it consumes the core without modifying it. --- Evaluation snapshot The current evaluation covers 68 matched model-dataset-topology configurations on BBH, GSM8K, and MMLU-Pro. Against matched LangGraph runtime graphs, observed latency changed by topology: Topology Latency vs. LangGraph Tokens vs. LangGraph Accuracy delta --- ---: ---: ---: Single agent -4.0% -1.5% +0.0 pp Three-agent chain +1.0% +3.1% +2.5 pp Fan-in -8.0% +1.7% +3.0 pp Fan-out -9.3% -0.1% +2.9 pp Runtime controls produced the largest savings when they avoided unnecessary model calls. In the evaluated ablations, early stopping reduced tokens by 52.6% and latency by 45.0% ; the adaptive policy reduced tokens by 21.2% and latency by 18.5% . A separate local studio-path benchmark used six paired synthetic workflows. gMAS","default_branch":null,"files":null,"tree":[],"storefront":"/r/frontier-ai-next","claimed":false,"request_supported":{"post":"https://gitbuyer.com/r/frontier-ai-next/flowMAS/request-supported","requests":0},"note":"indexed from public GitHub; nothing is for sale on this page. Clone it from GitHub. Paid listings live at /search."}