# 1flowbase [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/taichuy/1flowbase  
**GitHub Stars:** 279  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/1flowbase

## Description
Self-hosted AI gateway and app platform whose runtime APIs are exposed to agents through a built-in MCP gateway.

## Claude Desktop Quick Installation
No confirmed install details yet, so no config is published. Follow the setup steps in the repository above.

## Documentation

## What 1flowbase MCP server does

The 1flowbase MCP server exposes the runtime APIs of 1flowbase to AI agents through its built-in MCP gateway. 1flowbase combines an AI gateway with an application runtime, so the MCP surface is not limited to model requests. It can provide agent access to gateway functions, workflows, stored conversation data, business APIs and data, and React-based application components where those runtime capabilities are available.

The underlying platform is API-first. Operations used by the graphical interface are implemented through APIs, and the MCP gateway makes those operations available to agents. This lets an agent work with the same runtime used by the human-facing application rather than relying on a separate integration layer.

## How it works

1flowbase sits between agents and a self-hosted runtime. The runtime can unify access to AI models, retain conversations in PostgreSQL, publish workflows as virtual models, and support application data and UI development. MCP provides the agent-facing entry point for these functions.

Tool discovery is progressive rather than presenting every available operation at once. Agents can use the `list`, `get`, and `call` pattern to discover available tools, inspect a selected tool, and invoke it. This approach is intended for runtimes with a broad API surface, including gateway, workflow, data, and application operations.

The platform also supports conversion among Anthropic Messages, OpenAI Chat Completions, and OpenAI Responses protocols. That protocol handling belongs to the 1flowbase runtime; the MCP gateway exposes runtime capabilities to agents independently of the provider protocol used by an application.

## Setup and configuration

The project supports Docker deployment through shell, PowerShell, and Windows CMD scripts. The Docker quick start is designed to start 1flowbase without manual environment setup; the deployment prompts guide the remaining configuration.

A source-based setup is also documented. It requires Node.js 24 or newer, pnpm 12.11.2, Rust 1.99 or newer with Cargo, and Docker with Docker Compose for middleware such as PostgreSQL. The documented development process clones the repository, optionally warms Rust build caches, and starts the services with the repository’s development script.

The README does not specify MCP client configuration fields or required environment-variable names. It also does not document a separate standalone MCP package; the MCP gateway is part of the self-hosted 1flowbase runtime.

## Tools and capabilities

The 1flowbase MCP server can expose API-backed runtime areas including:

- AI gateway and unified model access
- Workflows and composite or virtual models
- Conversation data retained by the runtime
- Business APIs and dynamically created business data
- React application capabilities
- Progressive MCP tool discovery through `list`, `get`, and `call`

The platform’s application foundation supports workflows, APIs, business data, and React code blocks. Which individual tools an agent sees depends on the capabilities exposed by the running 1flowbase instance.

## Limitations and notes

1flowbase stores full conversations in PostgreSQL. The README warns that this can create substantial data volume, so production deployments should plan storage capacity, retention, backups, and compliance. These operational considerations apply to the runtime that hosts the MCP gateway.

The project describes the MCP tool descriptions as an area for future improvement. Agents may therefore need additional context about available operations, particularly when using the gateway to build or manage applications. The README does not provide a fixed tool catalog, client compatibility matrix, authentication reference, or environment-variable list, so those details must be confirmed from the deployed instance and current project documentation.

_Full upstream README: https://allmcps.com/mcp/1flowbase/readme_