Connect using the documented environment settings
SeedRouter's Claude Code guide records configuration and multi-step tests with Claude Code 2.1.280 on September 27, 2026. Use it as the detailed reference, including its model-specific limits and test scope.
Create a SeedRouter API key, then configure your shell:
export ANTHROPIC_BASE_URL=https://api.seedrouter.ai
export ANTHROPIC_API_KEY=sk-your-seedrouter-key
claude --model claude-fable-5-1Keep the key private. Check any existing model aliases or environment settings before starting a session, so background tasks use the models you intend.
Choose and verify a model
Use the exact identifier from the model directory. The configuration guide covers Claude models; the Kimi coding guide covers a different model family and its own setup. Compatibility details should remain attached to the model and client version that were checked.
For a project test, ask the assistant to read a small file, make a bounded change and run an existing check. Inspect both the actual file diff and the command result. A successful chat response alone does not verify tool use or a complete coding task.
Understand session cost
Coding sessions can include system instructions, tool definitions, project context and several model calls. Check the current model rates and your usage history rather than estimating a session from the price of a short prompt.
Preserve the existing model's API documentation as the source for supported parameters. Claude Code manages local tools and project permissions; SeedRouter supplies model access. The historical test record is not a guarantee that every newer client release or every model combination behaves identically.