ccflare 🛡️

A multi-provider native proxy for Anthropic and OpenAI.

ccflare routes each provider by URL prefix, load-balances across multiple accounts, and keeps full request history, rate-limit state, and usage analytics without translating provider payloads.

Why ccflare?

• Native passthrough — Anthropic stays Anthropic, OpenAI stays OpenAI
• Multi-provider routing — route by /v1/{provider}/*
• Compatibility routes — route by /v1/ccflare/* with family-prefixed models
• Account failover — retry another account when one provider account is rate limited
• Built-in observability — dashboard, request history, analytics, logs, and health endpoints
• Flexible auth — API key and OAuth account support

Quick start

git clone https://github.com/snipeship/ccflare
cd ccflare
bun install

# Start the server + dashboard on http://localhost:8080
bun run start

# Or launch the TUI, which can also start the server
bun run ccflare

Verify the server is up:

curl http://localhost:8080/health

How routing works

ccflare proxies requests by provider prefix:

• http://localhost:8080/v1/anthropic/*
• http://localhost:8080/v1/openai/*
• http://localhost:8080/v1/ccflare/*

Examples:

• /v1/anthropic/v1/messages → https://api.anthropic.com/v1/messages
• /v1/openai/chat/completions → https://api.openai.com/v1/chat/completions
• /v1/openai/responses → https://api.openai.com/v1/responses

The /v1/{provider} prefix is stripped exactly once before forwarding upstream.

Compatibility routes keep the client-facing schema but select a provider family from the model prefix:

• openai/<model-id> → prefers codex, then openai
• anthropic/<model-id> → prefers claude-code, then anthropic

Examples:

• /v1/ccflare/openai/chat/completions with "model":"openai/gpt-5.4"
• /v1/ccflare/openai/responses with "model":"anthropic/claude-sonnet-4"
• /v1/ccflare/anthropic/messages with "model":"openai/gpt-4o-mini"

Account setup

API key accounts

Add accounts through the management API:

curl -X POST http://localhost:8080/api/accounts \
  -H "content-type: application/json" \
  -d '{
    "name": "anthropic-main",
    "provider": "anthropic",
    "auth_method": "api_key",
    "api_key": "sk-ant-..."
  }'

curl -X POST http://localhost:8080/api/accounts \
  -H "content-type: application/json" \
  -d '{
    "name": "openai-main",
    "provider": "openai",
    "auth_method": "api_key",
    "api_key": "sk-openai-..."
  }'

OAuth accounts

Use the CLI/TUI for interactive OAuth setup:

# Claude Code OAuth
bun run ccflare --add-account work --provider claude-code

# Codex OAuth
bun run ccflare --add-account codex --provider codex

The management API also exposes provider-specific auth endpoints:

• POST /api/auth/anthropic/init
• POST /api/auth/anthropic/complete
• POST /api/auth/openai/init
• POST /api/auth/openai/complete

Provider configuration

Anthropic clients

Point Anthropic SDKs or curl at the Anthropic-prefixed base URL:

export ANTHROPIC_BASE_URL=http://localhost:8080/v1/anthropic

OpenAI clients

Point OpenAI-compatible clients at the OpenAI-prefixed base URL:

export OPENAI_BASE_URL=http://localhost:8080/v1/openai

You can configure both providers at the same time and ccflare will keep account selection isolated per provider.

Example usage

Anthropic example