AI Gateway
Access models and published workflows through one gateway, with conversion between three supported AI API protocols.
- ✓OpenAI Chat Completions
- ✓OpenAI Responses
- ✓Anthropic Messages
Product capabilities
From AI Gateway to a complete application runtime.
Distribute AI capabilities, retain complete conversations, and build with workflows, APIs, business data, and React. Self-host the foundation on infrastructure you control.
Workflow & virtual models
Combine models, tools, APIs, and business logic, then publish the workflow as a virtual model. A worker model can also become a tool for your primary model.
- ✓Visual workflow composition
- ✓Workflow published as a model
- ✓Model-as-tool delegation
Complete conversation data
Store the complete AI conversations that pass through your gateway in PostgreSQL. Use that history to analyze usage and improve prompts, routing, and model composition.
- ✓Complete AI sessions
- ✓Model and tool execution logs
- ✓Token and duration metrics
API-first backend
Runtime operations are built on APIs. Create business tables in PostgreSQL and manage application data alongside the AI conversations that inform it.
- ✓API-based runtime operations
- ✓Dynamic business tables
- ✓AI data and business data together
MCP Gateway
Expose APIs and MCP tools to Agents through progressive discovery. Find available tools, load the definition you need, then call the tool for the task.
- ✓list: discover available tools
- ✓get: retrieve a tool definition
- ✓call: execute the selected tool
React Blocks
Build application interfaces with React code blocks. Use the React ecosystem to connect your UI with the APIs, workflows, and data behind it.
- ✓Editable React code
- ✓React ecosystem components
- ✓Business application interfaces
Compose capabilities. Inspect how they run.
The workflow editor, API publishing, execution logs, and token dashboard provide a view into your AI runtime.
Self-host the runtime. Build from there.
Open source under Apache 2.0. Application templates and improved Agent building without source-code context are in progress. For complex applications today, keep your coding Agent in the 1flowbase project context.
View source ↗