Skip to main content
Claude Code talks to the Anthropic API, so you can point it at Relay and get usage tracking and full payload capture for every Claude Code session — without changing how you log in. Your existing Claude subscription or API key keeps working; Relay simply sits in the middle, observing and forwarding.

​
How it works

This uses proxy mode. Claude Code keeps authenticating to Anthropic with its own credentials — those flow through Relay untouched in the Authorization header. Relay identifies the traffic by your relay key, sent in X-WR-API-Key, and forwards the request verbatim to Anthropic.
Claude Code routed through Relay in proxy mode
Because Relay never substitutes the credential, your Claude Code auth (OAuth login or ANTHROPIC_API_KEY) is unchanged — Relay only adds observability.

​
Prerequisites

  1. Proxy mode enabled on your relay: 
    curl -X PUT https://your-relay/settings/proxy-mode \
  -H "Authorization: Bearer $RELAY_ADMIN_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"enabled": true}'
  2. A relay key whose policy grants the Claude models Claude Code uses (Sonnet, Opus, Haiku). In authenticated proxy mode without an explicit host header, Relay reads the model from the request and resolves the upstream host from this key’s policy — so the models must be granted.
  3. (Optional) Payload logging turned on, if you want to capture the full request/response bodies — see Usage & Logging.

​
Setup

Set three things via environment variables: the relay’s Anthropic base URL, and two custom headers (X-WR-API-Key and X-WR-Proxy-Mode). Claude Code’s own login stays as-is and flows through as the upstream credential.

​
Option A — a wrapper command

Drop a small script on your PATH (e.g. wcc) so you opt in per-invocation: 
#!/usr/bin/env bash
# Claude Code routed through Wyolet Relay.
export ANTHROPIC_BASE_URL="https://your-relay/anthropic"
export ANTHROPIC_CUSTOM_HEADERS=$'X-WR-API-Key: sk-wr-your-relay-key\nX-WR-Proxy-Mode: Proxy'
exec claude "$@"
Run wcc instead of claude when you want the session tracked; plain claude stays direct.

​
Option B — Claude Code settings

To route every session, set the same values under env in ~/.claude/settings.json (or a project-level .claude/settings.json): 
{
  "env": {
    "ANTHROPIC_BASE_URL": "https://your-relay/anthropic",
    "ANTHROPIC_CUSTOM_HEADERS": "X-WR-API-Key: sk-wr-your-relay-key\nX-WR-Proxy-Mode: Proxy"
  }
}
ANTHROPIC_BASE_URL is host + the /anthropic prefix; Claude Code appends /v1/messages itself. ANTHROPIC_CUSTOM_HEADERS is newline-separated Name: Value pairs.

​
What you get

Once a session runs through Relay:
  • Usage — every Claude Code request appears in GET /logs and the /usage/* endpoints, attributed to your relay key, with token counts and timing. Great for seeing what Claude Code actually costs.
  • Payloads — with payload logging enabled, the full prompts and completions are captured and readable at GET /logs/{request_id} — exactly what Claude Code sent and what the model returned.
See Usage & Logging for reading it back.
Keep your relay key out of shared dotfiles and version control — treat the wrapper script or settings.json as secret-bearing. Payload capture records prompt and completion text; scope it to the keys you mean to debug.

​
Notes

  • Setting ANTHROPIC_BASE_URL disables Claude Code’s MCP tool search by default; re-enable it with ENABLE_TOOL_SEARCH=true if you use it.
  • Proxy mode does no translation — Claude Code’s native Anthropic requests are forwarded byte-for-byte.