1flowbase MCP server
Commit Nov 15, 2026Lets an AI client access an AI gateway, workflows, dynamic business data, and React app backends through MCP.
Also known as taichuy/1flowbase
- Stars
- 293
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 15, 2026
- License
- Apache-2.0
What is 1flowbase MCP?
1flowbase lets an AI client access an AI gateway, workflows, dynamic business data, and React app backends through MCP.
Summary generated from the project's README
About 1flowbase MCP server
Connecting AI agents to a complete application backend is what 1flowbase enables through its MCP gateway. This self-hosted system serves as an AI gateway and application runtime that allows developers to unify model access while retaining conversation data in PostgreSQL. By exposing every runtime operation via MCP, agents can manage workflows, business data, and React-based application backends.
Users deploy the system using Docker or a Git-based setup requiring Node.js, pnpm, and a Rust toolchain. The architecture supports converting between Anthropic Messages and OpenAI Chat Completions protocols, ensuring the application layer remains stable even when switching providers. Instead of overwhelming the context, the gateway uses progressive list, get, and call tool discovery to help agents understand and manage the environment. This setup is designed for those building complex agentic systems that require dynamic business tables and native React code blocks rather than fixed drag-and-drop interfaces.
What you can do with it
- 1Manage business data via agentsAn AI agent can interact with dynamic business tables and CRUD APIs to manage application data.
- 2Execute complex multi-model workflowsDevelopers can combine high-capability planning models with low-cost execution models into a single virtual model tool.
- 3Build custom React interfacesUsers can create application UIs using the full React ecosystem instead of restricted visual editors.
- 4Analyze conversation assetsThe system stores full AI conversations to facilitate cost analysis and model quality evaluation.
Written from the project's README. Every use case is checked against a quote from it.
Automated checks
Ran Nov 14, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachabletaichuy/1flowbase is public and reachablegithub.com/taichuy/1flowbase ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)Apache-2.0spdx.org/licenses/Apache-2.0.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-13github.com/taichuy/1flowbase ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHub1flowbase
From AI Gateway to a Complete Application System.
1flowbase is a low-code native backend foundation built on an AI gateway. Beyond core AI gateway features, it lets you build your own application system on top of the AI chat data stored in the gateway.
An Application Foundation Designed for Agents
1flowbase is API-first by design. If a traditional GUI can take over the entire system through APIs, then turning those APIs into MCP lets an Agent take over everything as well.
Technical Architecture
1flowbase is a self-hosted AI Gateway and AI application runtime.
Start by unifying model access, retain conversation data in the same runtime, and keep building Workflows, APIs, business data, and React applications on top of it. Every capability is API-based and can be exposed to Agents through MCP to understand, build, and manage.
Dispatch intelligence, retain data, build applications.
Philosophy
- The gateway is not the end: Full AI conversations passing through the gateway are stored in PostgreSQL, becoming a reusable data asset for cost analysis, model quality evaluation, prompt / routing optimization, and business system development.
- A "model" can be a Workflow: High-capability models plan, judge, and audit; low-cost models execute. The combination is published as a single virtual model, and a model can itself be used as a Tool by another model.
- Three protocols, one entry point: Conversion between Anthropic Messages, OpenAI Chat Completions, and OpenAI Responses, so your application layer doesn't change when providers do.
- Designed for Agents: Every runtime operation available in the GUI is built on APIs, which are exposed to Agents via MCP. The MCP Gateway uses progressive
list / get / calltool discovery instead of stuffing every Tool into the context at once. - Low-code efficiency + native React freedom: Business tables can be created dynamically; UIs are written as React code blocks with access to the entire React ecosystem, not a closed drag-and-drop editor.
Human → GUI ─┐
├→ API → AI Gateway / Workflow / Data / React UI → AI Application
Agent → MCP ─┘
Quick Start
Docker
Deployed with Docker — no manual environment setup required.
Linux / macOS
curl -fsSL https://raw.githubusercontent.com/taichuy/1flowbase/main/scripts/shell/docker-deploy.sh | sh
Windows PowerShell
irm https://raw.githubusercontent.com/taichuy/1flowbase/main/scripts/powershell/docker-deploy.ps1 | iex
Windows CMD
powershell -NoProfile -ExecutionPolicy Bypass -Command "irm https://raw.githubusercontent.com/taichuy/1flowbase/main/scripts/powershell/docker-deploy.ps1 | iex"
Follow the prompts to finish configuration and start 1flowbase. Storing full conversations can generate large data volumes; for high-traffic production, plan storage capacity, retention, backup, and compliance in advance.
Git
Requirements
- Node.js
>=24.22.1 - pnpm
12.11.2(runcorepack enableandcorepack install --global pnpm@12.11.2) - Rust stable toolchain (Cargo), version
1.99or newer - Docker with Docker Compose (for middleware such as PostgreSQL)
Start
Run from the repository root:
git clone https://github.com/taichuy/1flowbase.git
cd 1flowbase
Before the first source build on Linux with a user systemd manager, apply the development resource profile and put its user-level wrappers on PATH:
bash scripts/shell/apply-resource-limits.sh
export PATH="$HOME/.local/bin:$PATH"
Rust normally uses half the available CPU and memory, targeting at least 2 logical CPUs and 8 GiB when capacity permits. Smaller machines may exceed the half-machine share; memory still leaves OS/service headroom. These are build budget targets, not verified minimum hardware requirements. Explicit local settings remain supported. Windows/macOS use the same CPU policy; Linux systemd memory limits do not apply there. See resource configuration for limits and overrides.
For first-time development, precompile the Rust test targets, then start the services:
node scripts/node/reset-rust-cache/cli.js --warmup-only
node scripts/node/dev-up.js
--warmup-only fetches locked dependencies and compiles build/test targets in sequence
without deleting existing caches or stopping services; it does not execute tests.
For subsequent daily starts, run dev-up directly to reuse the caches.
dev-up starts the Docker middleware, installs frontend dependencies, builds and runs the backend, and starts the frontend. Other common commands:
node scripts/node/dev-up.js status
node scripts/node/dev-up.js stop
node scripts/node/dev-up.js restart --frontend-only
node scripts/node/dev-up.js restart --backend-only
Logs are written to tmp/logs/. For more scripts, see scripts/README.md.
Comparison
| Project | AI Gateway | Composite Models / Workflow | Full Conversation Data | Dynamic Business Data | MCP | React Apps |
|---|---|---|---|---|---|---|
| 1flowbase | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
| LiteLLM | ✅ | — | Partial | — | — | — |
| OpenRouter | ✅ | — | Platform-hosted | — | — | — |
| Dify | Partial | ✅ | ✅ | — | ✅ | Fixed app types |
| Supabase | — | — | — | ✅ | Ecosystem | Build your own |
| n8n | — | ✅ | — | — | Ecosystem | — |
Roadmap
- Out-of-the-box application templates (AI Gateway, multi-model router, enterprise AI Gateway, AI education platform, Agent application backend)
- Better official MCP Tool descriptions so Coding Agents like Codex can build applications without reading the source code
- Lifecycle management for AI conversation data
Tutorials
- Make GLM-5.2 See Images in Claude Code with 1flowbase
- Fusion-Style Workflow: Publish a Multi-Model Review Panel as an Observable Virtual Model
- 1flowbase Wiki
Repository Layout
web/ Frontend root, powered by pnpm + Turbo
api/ Rust backend workspace
api/apps/ Backend service entry points
api/crates/ Shared backend crates
api/plugins/ Plugin source workspace, HostExtension manifests and templates
docker/ Local middleware orchestration and self-hosted service stack
scripts/ Repository-level development, testing, verification, and debugging scripts
Contributing
Community contributions are very welcome. Before submitting a Pull Request, please run the following verification script:
node scripts/node/verify.js repo
Development guidelines:
Friends
- Linux.do - Learn AI, on L Station.
- openai - Thanks to the early OpenAI Developer Program for supporting this project.
- Aionui - Remotely control AI from your phone.
- OfficeCLI - Office suite designed for AI Agents.
- deepseek-pp - DeepSeek web chat browser extension.
- MuseAI - Local AI companion, text adventure, and interactive fiction app.
- FrontAgent - AI Agent system designed for frontend engineering.
- RedBox - Localized AI creation workspace for Xiaohongshu creators.
License
This project is licensed under Apache-2.0.
Contributors
Star History
If you want Agents to build and operate self-hosted applications across AI, MCP, application backends, and React interfaces — give 1flowbase a Star.
1flowbase MCP questions
What is 1flowbase?
1flowbase is a self-hosted AI gateway and AI application runtime that unifies model access and lets you build workflows, APIs, business data, and React applications on top of it.
How does 1flowbase expose capabilities to agents?
Every runtime operation is API-based, and the APIs are exposed to agents through MCP. The MCP Gateway uses progressive list/get/call tool discovery.
Which model provider protocols does 1flowbase support?
It converts between Anthropic Messages, OpenAI Chat Completions, and OpenAI Responses, so the application layer does not change when providers do.
Where are AI conversations stored?
Full AI conversations passing through the gateway are stored in PostgreSQL and become reusable data for cost analysis, model evaluation, and business system development.
Can a workflow be used as a model?
Yes. Models can be combined into a published virtual model, and a model can itself be used as a tool by another model.