Install and verify

This page takes you from a key to a first chat completion. You are done when a completion comes back with text in it. Pick TypeScript or Python. The CLI is not required for any of this.

Four steps#

1. Get an API key

   Open conifer.build/console#/keys and mint a key. You can do this before entering a card. A key is free to hold and spends nothing until it is used. The key is shown once, so save it now.

   macOS / LinuxWindows

   terminal

   export CONIFER_API_KEY='sk-conifer-…'

   Your first call needs a balance on the account. Set up pay as you go at Account & billing. Each model is priced per token as published in the catalog.

2. Install a client

   There are two supported choices. If you are unsure, take the first. The official OpenAI package is the ordinary client from npm or PyPI, the same package you would install to call OpenAI directly. Conifer speaks the OpenAI wire, so the package works unmodified. For the Anthropic wire, install npm install @anthropic-ai/sdk or pip install anthropic instead.

   npmpip

   terminal

   npm install openai

   The second choice is our own open-source client at ConiferKit/use-conifer. It gives you the exact cost of every call and a hard spend ceiling. It has one unscoped name on both registries. Keep the [tls] extra on macOS. See The Conifer SDK.

   npmpip

   terminal

   npm i conifer-sdk

3. Point the client at Conifer

   Either pass the base URL and key in code, or set the two variables the OpenAI client already reads and leave the source unchanged.

   terminal

   export OPENAI_BASE_URL=https://api.conifer.build/v1
   export OPENAI_API_KEY=$CONIFER_API_KEY

4. Make one call

   The TypeScript sample is an App Router handler. Paste it into app/api/verify/route.ts. The Python sample is a script you run directly.

   TypeScriptPythoncurl

   app/api/verify/route.ts

   import OpenAI from "openai";
   
   export function conifer() {
     const apiKey = process.env.CONIFER_API_KEY;
     if (!apiKey) {
       throw new Error("CONIFER_API_KEY is missing. Set it before calling the gateway.");
     }
     return new OpenAI({
       baseURL: "https://api.conifer.build/v1",
       apiKey,
     });
   }
   
   export async function POST(req: Request) {
     const res = await conifer().chat.completions.create({
       model: "claude-haiku-4-5",
       max_tokens: 64,
       messages: [{ role: "user", content: "Say hello from Conifer and nothing else." }],
     });
   
     return Response.json({ content: res.choices[0].message.content });
   }

If it did not work#

• 401, invalid_api_key. The key is missing, mistyped, or revoked. Check that the variable reached the process.
• 402, insufficient_quota. The key is valid but the account has no balance. Set up billing. Nothing was charged.
• 404, model_not_found. That id is not in the catalog. The catalog is public and needs no key.
  terminal

  curl -s https://api.conifer.build/v1/catalog

  GET /v1/models is the same list scoped to your key. See the catalog.

Every error name is on Errors.

Optional: the MCP server#

The SDK package also ships an MCP server. With it, an agent that is already driving your editor can ask any model in the catalog mid-task and see what every call cost. There is no build step. The package is conifer-sdk, unscoped.

{
  "mcpServers": {
    "conifer": {
      "command": "npx",
      "args": ["-y", "conifer-sdk", "conifer-mcp"],
      "env": { "CONIFER_API_KEY": "sk-conifer-…" }
    }
  }
}

Claude Code#

claude mcp add conifer --env CONIFER_API_KEY=sk-conifer-… -- npx -y conifer-sdk conifer-mcp

Claude Desktop#

Claude Desktop is a different program from Claude Code, with a different config file. Open Settings, then Developer, then Edit Config to open claude_desktop_config.json. Afterwards quit, not just close the window, and reopen.

{
  "mcpServers": {
    "conifer": {
      "command": "/absolute/path/to/npx",
      "args": ["-y", "conifer-sdk", "conifer-mcp"],
      "env": { "CONIFER_API_KEY": "sk-conifer-…" }
    }
  }
}

Codex#

[mcp_servers.conifer]
command = "npx"
args = ["-y", "conifer-sdk", "conifer-mcp"]
env = { "CONIFER_API_KEY" = "sk-conifer-…" }

Cursor#

{
  "mcpServers": {
    "conifer": {
      "command": "npx",
      "args": ["-y", "conifer-sdk", "conifer-mcp"],
      "env": { "CONIFER_API_KEY": "sk-conifer-…" }
    }
  }
}

VS Code#

The inputs block makes VS Code prompt for the key instead of committing it.

{
  "inputs": [
    {
      "type": "promptString",
      "id": "conifer-key",
      "description": "Conifer API key",
      "password": true
    }
  ],
  "servers": {
    "conifer": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "conifer-sdk", "conifer-mcp"],
      "env": { "CONIFER_API_KEY": "${input:conifer-key}" }
    }
  }
}

What the agent gets#

Each model’s call in a comparison is its own billed turn. max_cost_nanousd caps each turn, not the total.