Provider Catalog

This page is the provider-first reference for cliproxyapi++: what each provider block is for, how to configure it, and when to use it.

Provider Groups

Group	Primary Use	Config Blocks
Direct APIs	Lowest translation overhead, direct vendor features	claude-api-key, gemini-api-key, codex-api-key, deepseek, groq, mistral
Aggregators	Broad model inventory under one account	openrouter, together, fireworks, novita, siliconflow, openai-compatibility
OAuth / Session Flows	IDE-style account login and managed refresh	kiro, cursor, minimax, roo, kilo, ampcode
Compatibility Endpoints	OpenAI-shaped upstream endpoints	openai-compatibility, vertex-api-key

Minimal Provider Patterns

1) Direct vendor key

claude-api-key:
  - api-key: "sk-ant-..."
    prefix: "claude-prod"

2) Aggregator provider

openrouter:
  - api-key: "sk-or-v1-..."
    base-url: "https://openrouter.ai/api/v1"
    prefix: "or"

3) OpenAI-compatible provider registry

openai-compatibility:
  - name: "openrouter"
    prefix: "or"
    base-url: "https://openrouter.ai/api/v1"
    api-key-entries:
      - api-key: "sk-or-v1-..."

3b) Orchids reverse proxy (OpenAI-compatible)

openai-compatibility:
  - name: "orchids"
    prefix: "orchids"
    base-url: "https://<your-orchids-endpoint>/v1"
    api-key-entries:
      - api-key: "<orchids-api-key>"

Use this when Orchids is exposed as an OpenAI-shaped /v1 endpoint and you want prefix-isolated routing (orchids/<model>).

4) OAuth/session provider

kiro:
  - token-file: "~/.aws/sso/cache/kiro-auth-token.json"

5) Kilo free-model endpoint (OpenRouter-compatible)

kilo:
  - api-key: "anonymous"
    base-url: "https://api.kilo.ai/api/openrouter"

Prefixing and Model Scope

• prefix isolates traffic per credential/provider (for example prod/claude-3-5-sonnet).
• force-model-prefix: true enforces explicit provider routing.
• models with alias gives client-stable names while preserving upstream model IDs.
• excluded-models prevents unsafe or expensive models from appearing in /v1/models.

Provider Selection Guide

Goal	Recommended Pattern
Predictable latency	Prefer direct providers (claude-api-key, gemini-api-key, codex-api-key)
Broad fallback options	Add one aggregator (openrouter or openai-compatibility)
Team/workload isolation	Use provider prefix and force-model-prefix: true
Zero-downtime auth	Use OAuth/session providers with token file refresh (kiro, cursor, minimax)
Lowest ops friction	Standardize all non-direct integrations under openai-compatibility

Validation Checklist

1. curl -sS http://localhost:8317/v1/models -H "Authorization: Bearer <api-key>" | jq '.data[].id'
2. Ensure required prefixes are visible in returned model IDs.
3. Issue one request per critical model path.
4. Check metrics: curl -sS http://localhost:8317/v1/metrics/providers | jq.
5. Confirm no sustained 429 or 401/403 on target providers.

Related Docs

• Provider Usage
• Provider Operations
• Routing and Models Reference
• OpenAI-Compatible API