Proxy
One OpenAI-style endpoint for chat completions, embeddings, models, and streaming across multiple providers.
flow llm-router
by FlowAPI
Local-First
OpenAI-Compatible
LLM Router
One endpoint, one dashboard, and full control over routing, credentials, security, and observability for your team's model traffic.
Install
pip install flow-llm-router
Base install for the local-first OpenAI-compatible gateway.
Local SQLiteMulti-providerSmart routing
Open the repository →One endpoint for models,
routing, and local observability
Preview
Agents, apps, and operators on one gateway
LOCAL GATEWAY
$ flow-router start
Serving FlowGate on http://127.0.0.1:7798
Dashboard ready at http://127.0.0.1:7798
Proxy base URL: http://127.0.0.1:7798/v1
$ curl http://127.0.0.1:7798/v1/models
{
"data": ["gpt-4o", "claude-sonnet", "gemini-2.5-pro"]
}
Control
Encrypted provider vault, caller tokens, IP allowlisting, and local SQLite logs keep the control plane in your hands.
Router
Pass-through, complexity scoring, and classifier routing all fit behind the same gateway without changing your app code.
Why it fits
Built for teams who want local control without giving up modern routing.
FlowGate sits between your apps and the model providers, keeping the operator surface close to your stack while still exposing a drop-in OpenAI-style interface for existing SDKs and agent code.
Quick start
1. flow-router start
2. Add provider keys in the dashboard
3. Point your OpenAI SDK at /v1
OpenAI-compatible base URL
Local observability in SQLite
Encrypted provider key vault
Built-in dashboard and logs
Rule-based and classifier routing
Multi-provider model catalog sync
Flow llm-router
Self-host the gateway layer your AI stack has been missing.
Route across OpenAI, Anthropic, Gemini, DeepSeek, Qwen, Groq, and custom OpenAI-compatible backends from one place, then keep logs, routing policy, and credentials under your own roof.