AI infrastructure · 2026
ModelPort
One private port for every model provider.
BUILT BYAGLocal first
Product surfaceIllustrated from the working application
Overview
24h7d30d
Total tokens4.82M
↑ 12% vs last week
Estimated cost$18.40
across 5 providers
Top modelsonnet-5.5
via Anthropic
Average latency842 ms
↓ 9% vs last week
Token usage over time
Showing total token usage for the selected date range
Recent requests
ClientProviderModelTokensStatus
Claude CodeAnthropicsonnet-5.512,408200
CursorOpenAI → Anthropicopus-5.58,190200
openai-pythonOpenRoutergemini-3-pro3,276fallback
CodexOllamaqwen3-coder1,944200
Claude CodeAnthropichaiku-5.5906200
anthropic-sdkGemini → OpenAIgemini-3-flash2,512200
01 / Context
The challenge
Developer tools often need direct access to provider keys and APIs. That couples every client to a vendor, spreads credentials across machines, and makes cost and request behavior difficult to understand.
The response
ModelPort creates a local control plane between clients and AI providers. It translates OpenAI-compatible requests into Anthropic-compatible requests and vice versa, secures credentials, routes traffic, and makes model access observable from one dashboard.
02 / What it does
Built around the
real work.
- 01
Bidirectional OpenAI and Anthropic request translation
- 02
OpenAI and Anthropic-compatible client endpoints
- 03
Local credential vault and bearer-token access
- 04
Provider routing, usage, cost, and request inspection
03 / Built with
FastAPIReactSQLiteLocal TLS
04 / Read more
One port for every modelWhy I put a local control plane between my AI tools and their providers, and what it takes to translate between OpenAI and Anthropic request formats in both directions.