Integrations

Drop GammaInfra's smart router into the tools you already use. Anything that accepts a custom base_url works as a drop-in — every prompt is classified by task and dispatched to the best-fit model across every major LLM. These pages cover the most common tools step by step.

Model Context Protocol

MCP server→
Call smart routing as a native tool from any MCP host — Claude Code, Claude Desktop, Cursor, Cline, Continue. npx -y @gammainfra/mcp-server; per-request cost and routing decision returned inline.

Code editors & agents

Cursor→
Cost-aware routing for AI-first code editing. See X-GammaInfra-Cost-USD on every completion, get automatic fallback when your provider rate-limits.
Cline→
Multi-provider fallback for agentic loops. When one provider throttles mid-loop, GammaInfra cascades to the next — your 20-tool-call task doesn't die at step 14.
Aider→
Cost-per-edit visibility from the response header — see what each iteration spent. CLI pair-programmer.
Continue→
One config entry covers the whole catalog. Per-role models for chat, autocomplete, edit.
Roo Code→
Cline fork with extended agent features. Same setup as Cline; same fallback story, more value on long auto-approve sessions.
Zed→
Fast Rust-based editor with native AI assistant. One settings.json change adds GammaInfra.

Frameworks

LangChain & LlamaIndex→
Drop-in for any OpenAI-shaped client. Your existing chain code unchanged — you just see costs and fallbacks. A/B compare providers without rewriting chains.
Vercel AI SDK→
Use GammaInfra via the @ai-sdk/openai provider's custom baseURL. Edge-runtime compatible. Streaming, tools, structured output.

Self-hosted UIs

Open WebUI→
Self-hosted ChatGPT-style UI against every major LLM with one API key.
LibreChat→
Add GammaInfra as a custom endpoint in librechat.yaml. Full model catalog, optional per-model token pricing.

What if my tool isn't listed?

Any tool that supports a custom OpenAI base URL works. The pattern is always the same — set the base URL to GammaInfra's, paste your GammaInfra API key, pick a model name. If the tool defaults to https://api.openai.com/v1, change it to:

https://api.gammainfra.com/v1

That's it. Your chosen model name can be any of these:

Full catalog at GET /v1/models.

What you get

Once a tool is routed through GammaInfra, every response carries:

Ready to try it?

Get a GammaInfra API key →

$3 free trial credit on signup, $10 minimum top-up. Pass-through provider token rates plus 3% top-up fee during the launch window (5% after 2026-06-23). BYOK alternative at 1–2% per request.

Frequently asked questions

Which IDE, agent, and chat tools work with GammaInfra?
Every tool that accepts a custom OpenAI-compatible base URL — which is essentially every modern AI dev tool. Direct setup guides exist for Cursor, Cline, Continue, aider, LangChain, Roo Code, Zed, Open WebUI, LibreChat, and the Vercel AI SDK. Any tool not listed but supporting OpenAI-compat works the same way: replace api.openai.com/v1 with api.gammainfra.com/v1.
Do all integrations use the same API key?
They can. One sk-gammainfra-... key works across every tool that points at the gateway. For per-tool attribution and isolated rate-limit headroom, issue separate keys per tool from the GammaInfra dashboard — same account, multiple keys.
Does smart routing work for every tool listed?
Yes. Use the model name gammainfra/auto in any tool's configuration and the router picks the best-fit model per request. Or pin a specific direct-pin model like anthropic/claude-opus-4-7 in tools that benefit from a fixed quality level. The wire format is the same so there's no per-tool routing logic to worry about.
How do I add an integration that isn't listed?
If your tool exposes an OpenAI-compatible custom-endpoint setting (Base URL + API Key + Model Name), the setup is identical to any of the listed integrations. Set Base URL to https://api.gammainfra.com/v1, API Key to your sk-gammainfra-... value, and Model to gammainfra/auto or any specific model. If your tool supports only the native API of a specific vendor (no custom-endpoint setting), GammaInfra can't redirect it.
Can I set per-tool rate limits?
The gateway enforces 240 rpm per API key. The cleanest per-tool pattern is to issue one key per tool from the dashboard — each tool gets its own rate-limit headroom and the dashboard's per-key filter gives clean attribution. For very heavy tools, ask support to raise the cap on that specific key.