v1 · Beta

One API for AI model intelligence.

Compare provider deployments using normalized pricing, capabilities, lifecycle, constraints, and evidence. Your provider or gateway still executes model traffic.

Why ALLM

A model name is only one layer of identity. A model release describes what a lab built; a deployment describes how a provider exposes it in a region through a specific API surface. Prices, limits, and capabilities attach to the deployment.

API basics

Base URL
https://api.use-allm.com
Media type
application/json for success; application/problem+json for errors.
Authentication
Bearer API key. Public catalog reads may be anonymous at lower limits.
Versioning
The major API version is part of every path, for example /v1/deployments.

Data model

model family → model release → deployment → rate card
deployment → lifecycle milestones → evidence
deployment → capability assertion → evidence
catalog release → immutable snapshot → lockfile

Your first intelligence query

Install the TypeScript SDK, create a free API key, then query deployments by the capabilities your product needs.

TypeScript
import { ALLM } from "@allm/sdk";

const allm = new ALLM({ apiKey: process.env.ALLM_API_KEY });
const deployments = await allm.deployments.list({
  model: "anthropic/claude-haiku-4.5",
  capability: "tool_calling",
});
JSON response
{
  "data": [
    {
      "id": "dep_anthropic_claude_haiku_4_5_messages_global",
      "model_id": "anthropic/claude-haiku-4.5",
      "provider_id": "anthropic",
      "region": "global",
      "capabilities": {
        "tool_calling": {
          "status": "supported",
          "verification_status": "official_declared"
        }
      }
    }
  ],
  "meta": { "total": 3, "limit": 100, "offset": 0, "has_more": false }
}

Persist decision metadata

When a result affects production routing or cost, store the deployment ID and x-allm-catalog-release together.

Where to go next