One inference platform.
Every model, every provider.
Stop wiring up a dozen providers. Greyroom handles model routing, provider management, and tool execution so you can just build.
What is Greyroom?
Greyroom is Floatpoint's shared inference platform. Instead of every product wiring up its own connections to a dozen different AI providers, they all go through Greyroom.
It picks the right model and provider for each request, handles the call, and returns a clean, structured response. Services like Observe AMS, CRQ, and everything else we build all run through the same layer.
Greyroom isn't generating AI or making decisions. It's infrastructure that routes inference requests to the best available model and gives you predictable results every time.
Core Capabilities
One platform between your products and a growing list of AI providers. Greyroom handles the complexity so your teams don't have to.
Smart Routing
Picks the right model and provider for each request based on the task, cost, and latency needs. No hardcoded provider logic in your product code.
Unified API
One interface for every provider. Swap models, switch providers, or add new ones without touching a line of product code.
Tool Execution
Structured tool calling with predictable results. Every invocation gets validated and sandboxed, so products can trust what comes back.
Shared Primitives
Common building blocks for classification, extraction, summarization, and scoring. Solved once, reused across every product.
Built-in Guardrails
Every request runs through the same safety checks and content policies. No product ships without them.
Usage Tracking
See token usage, costs, and latency across every provider and product from one place. No more surprise bills.
How It Works
Greyroom sits between your products and the providers. Products send inference requests, Greyroom routes them to the right model and sends back structured responses.
Design Principle
Greyroom is infrastructure, not intelligence. It routes inference requests to the best available model, manages provider connections, and returns structured results. The models do the thinking. Greyroom just makes sure the right one gets the job.
Work With Us
Greyroom is used internally at Floatpoint and doesn't have a public API. But if you're a developer who wants to build on a managed inference layer, we're open to working together on custom implementations.
Get in touch.
projects@greyroom.dev