Skip to main content

Claude Code

Claude Code routes through Voxvey's Messages API surface.

POST https://api.voxvey.com/v1/messages
GET https://api.voxvey.com/v1/models

Claude Code sends inference requests to /v1/messages when ANTHROPIC_BASE_URL points at a gateway that exposes the Messages API format.

1. Set the token

export VOXVEY_TOKEN="<access-token>"

2. Point Claude Code at Voxvey

export ANTHROPIC_BASE_URL="https://api.voxvey.com"
export ANTHROPIC_AUTH_TOKEN="$VOXVEY_TOKEN"
export CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1
claude

ANTHROPIC_AUTH_TOKEN sends the Voxvey token as a bearer token. Use this instead of ANTHROPIC_API_KEY for Voxvey.

3. Select a model

When gateway model discovery is enabled, Claude Code queries https://api.voxvey.com/v1/models at startup and adds compatible returned models to /model.

/model

If discovery is disabled or you want to pin a model explicitly, set a custom model option before starting Claude Code:

export ANTHROPIC_CUSTOM_MODEL_OPTION="<messages-model-id>"
export ANTHROPIC_CUSTOM_MODEL_OPTION_NAME="Voxvey Messages"
claude --model "<messages-model-id>"

Replace <messages-model-id> with a model ID returned by Voxvey for the Messages API. Claude Code only auto-adds discovered model IDs that begin with claude or anthropic, so use ANTHROPIC_CUSTOM_MODEL_OPTION for any other ID shape.

4. Verify

Inside Claude Code:

/status
/model

Then send:

Reply with the exact word ready.

If requests fail with 404s for /v1/messages, verify that the Messages route is deployed. If /model does not show Voxvey models, keep the custom model option set or check the /v1/models response.