Claude Code and IDE tools

Coding tools talk to Jantaris through the two compatible endpoints. Claude Code uses /v1/messages (Anthropic format). Cursor, Cline, Continue and most other tools use /v1/chat/completions (OpenAI format). In both cases you change the base URL and the key, nothing else.

Create a separate API key for your coding tools and give it a monthly spend limit in the dashboard. Agents can send many requests in a long session, and the limit caps the damage of a runaway loop.

Claude Code

macOS and Linux

export ANTHROPIC_BASE_URL=https://jantaris.com
export ANTHROPIC_AUTH_TOKEN=sk-jt-...
export ANTHROPIC_MODEL=claude-opus-5.5   # optional
claude

Add the lines to ~/.bashrc or ~/.zshrc to keep them across sessions.

Windows PowerShell

$env:ANTHROPIC_BASE_URL = "https://jantaris.com"
$env:ANTHROPIC_AUTH_TOKEN = "sk-jt-..."
$env:ANTHROPIC_MODEL = "claude-opus-5.5"   # optional
claude

To make them permanent for your user:

[Environment]::SetEnvironmentVariable("ANTHROPIC_BASE_URL", "https://jantaris.com", "User")
[Environment]::SetEnvironmentVariable("ANTHROPIC_AUTH_TOKEN", "sk-jt-...", "User")

Open a new terminal afterwards.

settings.json

Instead of shell variables you can put the configuration in Claude Code's settings file, ~/.claude/settings.json for all projects or .claude/settings.json inside one project:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://jantaris.com",
    "ANTHROPIC_AUTH_TOKEN": "sk-jt-...",
    "ANTHROPIC_MODEL": "claude-opus-5.5"
  }
}

Do not commit a project settings file that contains your key. Keep the key in the user-level file or in .claude/settings.local.json.

Choosing the model

You can also start a session with a specific model: claude --model claude-sonnet-5. Only Claude models are available on /v1/messages: Claude Opus 5.5 ($5.40 / $27.00 per 1M tokens), Claude Sonnet 5 ($3.24 / $16.20 per 1M tokens), Claude Fable 5.1 ($10.80 / $54.00 per 1M tokens).

Checking the setup

Run /status inside Claude Code and confirm the base URL points to Jantaris. If requests fail:

The 1M-token context mode is not available - Jantaris removes context-1m-* beta flags (and the [1m] model suffix), so every request runs in the model's standard context window.

To see how much a session cost, compare GET /v1/balance before and after.

Generate video and images from Claude Code (MCP)

Claude Code can also generate video, images and music. Connect the Jantaris MCP server with one command:

claude mcp add --transport http jantaris https://jantaris.com/mcp --header "Authorization: Bearer sk-jt-..."

Then ask, for example, "make a 5-second 9:16 Seedance 2.5 reel about coffee, show the cost first". Tool list, scopes (--scope user) and safety: MCP server.

Cursor

1. Open Settings, then Models.
2. In the OpenAI API key section, paste your sk-jt-... key.
3. Turn on Override OpenAI Base URL and enter https://jantaris.com/v1.
4. Add a custom model with a Jantaris id, for example claude-sonnet-5 or gpt-6-sol, and select it in chat.

Cline

In the Cline settings panel:

Continue

In ~/.continue/config.yaml:

models:
  - name: Jantaris Claude Sonnet 5
    provider: openai
    model: claude-sonnet-5
    apiBase: https://jantaris.com/v1
    apiKey: sk-jt-...
  - name: Jantaris DeepSeek
    provider: openai
    model: deepseek-v4.1-flash
    apiBase: https://jantaris.com/v1
    apiKey: sk-jt-...

Roo Code

In Roo Code settings choose API Provider: OpenAI Compatible and enter the same values as for Cline: Base URL https://jantaris.com/v1, API Key sk-jt-..., Model claude-sonnet-5 (or any other chat model).

OpenCode

In opencode.json (in the project or in ~/.config/opencode/):

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "jantaris": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Jantaris",
      "options": {
        "baseURL": "https://jantaris.com/v1",
        "apiKey": "{env:JANTARIS_API_KEY}"
      },
      "models": {
        "claude-sonnet-5": { "name": "Claude Sonnet 5" },
        "gpt-6-sol": { "name": "GPT-6 Sol" }
      }
    }
  }
}

Set the key in the JANTARIS_API_KEY variable. Add the MCP server for media in the same file, see MCP server.

Aider

macOS and Linux:

export OPENAI_API_BASE=https://jantaris.com/v1
export OPENAI_API_KEY=sk-jt-...
aider --model openai/claude-sonnet-5

Windows (then open a new terminal):

setx OPENAI_API_BASE https://jantaris.com/v1
setx OPENAI_API_KEY sk-jt-...
aider --model openai/claude-sonnet-5

The openai/ prefix tells Aider it is an OpenAI-compatible endpoint. After it goes the Jantaris model id.

Zed

In Zed's settings.json:

{
  "language_models": {
    "openai_compatible": {
      "Jantaris": {
        "api_url": "https://jantaris.com/v1",
        "available_models": [
          { "name": "claude-sonnet-5", "display_name": "Claude Sonnet 5", "max_tokens": 200000 }
        ]
      }
    }
  }
}

Enter the key in the provider settings of Zed's AI panel (not in settings.json). max_tokens tells Zed how much context it may send. 200000 is a safe value; check each model's window on its page in the catalog.

Other OpenAI-compatible tools

Most tools built on the OpenAI SDK read these environment variables:

export OPENAI_BASE_URL=https://jantaris.com/v1
export OPENAI_API_KEY=sk-jt-...

If a tool has a "custom endpoint", "OpenAI-compatible provider" or "base URL" field, the values are the same: https://jantaris.com/v1, your key and a model id from Models.

The full tool map, including Codex CLI, Claude Desktop and VS Code: Integrations.