1flowbase
Orchestrates complex multi-model AI workflows behind a single, observable virtual model endpoint for agent-native applications.
1flowbase is an open-source, full-stack platform designed for building agent-native applications. It functions as an AI Gateway and application backend, enabling developers to create and manage intricate multi-model workflows using a visual editor. By publishing these workflows as standard OpenAI-compatible or Claude-compatible APIs, it allows local AI agents to leverage advanced capabilities like vision models for screenshots, multi-model review panels, and robust output structuring, all while providing comprehensive observability for tracing model calls, tool callbacks, token usage, latency, and costs.
Key Features
01Visual workflow editor for multi-node AI orchestration
02Support for streaming responses and mounted LLM tools (e.g., vision models)
03Self-hosted platform with generated CRUD APIs and native React frontend blocks
04Publish workflows as OpenAI-compatible or Claude-compatible virtual model APIs
05Comprehensive execution logs with tool callback traces, token usage, and latency tracking
06224 GitHub stars
Use Cases
01Publish a Fusion-style multi-model reviewer panel for synthesizing results from multiple LLMs
02Add vision capabilities to text-first coding models (e.g., routing screenshots to a vision model)
03Build programmable upstream models for agents that include testing, linting, and review nodes