Paid Providers

Configure paid API providers securely in ClawRouter. All keys are stored locally -- never sent externally.

Version 1.0.18

Important: ClawRouter only supports standard Developer API Keys. It does NOT support web session tokens, OAuth logins, or consumer subscriptions (e.g., ChatGPT Plus or Claude Pro web credentials). You must generate an actual API Key from the provider's developer console.

Client configuration: Ready-to-paste setups for every client (OpenClaw, OpenCode, Claude Code, Codex CLI, Qwen Code, DeepSeek Harness, custom) live in the Client Setup section -- or open the provider's Prompt for AI dialog in the dashboard, which generates the config for you with your real proxy key and current model list included.

---

Perplexity

Perplexity offers two separate presets:

Perplexity (Sonar API)

Native search-augmented models: sonar, sonar-pro, sonar-reasoning-pro, sonar-deep-research

Perplexity Agent (Responses API)

Third-party frontier models with optional web search via the OpenAI Responses API format: perplexity/sonar, openai/gpt-5.4, anthropic/claude-sonnet-4-6, google/gemini-3.1-pro-preview

Live Model List: Perplexity has a real /v1/models endpoint -- Fetch Models retrieves the current catalog (45+ models, including Agent-API gateway ids like anthropic/..., openai/..., perplexity/glm-5.x, perplexity/kimi-k2.7-code) with per-model pricing. The seeded lists above are only the starting set.

Model lists change over time -- snapshot from August 2026. Official catalog: docs.perplexity.ai/docs/sonar/models

---

OpenAI

Seeded models: gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna, gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.2, gpt-5.1

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list. Official catalog: platform.openai.com/docs/models

Dashboard Setup

1. Quick Setup > OpenAI > Create Provider.
2. Add your OpenAI API key(s) in the API Keys tab.
3. Copy the Base URL.

---

Anthropic

Seeded models: claude-fable-5, claude-opus-5, claude-sonnet-5, claude-haiku-4-5-20251001

Note: Anthropic uses the anthropic-messages API format, not openai-completions. ClawRouter auto-fills the required anthropic-version: 2023-06-01 header when your client doesn't send it.

Live Model List: Anthropic has a real /v1/models endpoint -- Fetch Models retrieves the current catalog (paginated). A hardcoded list is used only as a fallback.

Model lists change over time -- snapshot from August 2026. Official catalog: platform.claude.com/docs/en/about-claude/models/overview

Dashboard Setup

1. Quick Setup > Anthropic > Create Provider.
2. Add your Anthropic API key(s) in the API Keys tab.
3. Copy the Base URL.

---

DeepSeek

Seeded models: deepseek-v4-pro, deepseek-v4-flash

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list. Official catalog: api-docs.deepseek.com/quick_start/pricing

---

xAI (Grok)

Seeded models: grok-4.6, grok-4.5, grok-4.3, grok-code-fast-1, grok-build-0.1

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list. Official catalog: docs.x.ai/docs/models

---

Mistral

Seeded models: mistral-large-latest, codestral-latest, mistral-medium-latest

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list. Official catalog: docs.mistral.ai/getting-started/models/models_overview

---

Z.AI

Z.AI offers three separate presets:

Z.AI API (General)

Z.AI Coding

Seeded models: glm-5.3, glm-5-turbo, glm-5v-turbo, glm-5.2, glm-5.1, glm-4.7

Z.AI Coding (China)

Seeded models: glm-5.3, glm-5-turbo, glm-5v-turbo, glm-5.2, glm-5.1, glm-5, glm-4.7

Model lists change over time -- snapshot from August 2026. The coding-plan gateway has no public model-list endpoint, so Fetch Models uses the curated list above; the general Z.AI API preset fetches live. Official docs: docs.z.ai

Business Error Codes

Z.AI returns a numeric-string business code in error.code that beats the HTTP status -- hard billing, window quotas, model gating, and rate limits all ride on 429. ClawRouter classifies by code:

Quota Visibility (GLM Coding Plan)

The Z.AI Coding presets have a Quota tab (see Provider Management > View Live Quota & Usage) backed by Z.AI's plan endpoints:

• GET {origin}/api/monitor/usage/quota/limit -- token windows (5-hour session, weekly -- the weekly window may be percentage-only and renders as %) with epoch-millisecond reset times, plus the monthly web-search count
• GET {origin}/api/biz/subscription/list -- plan name (best-effort)

Two traps handled automatically:

• A key with no GLM Coding Plan returns a "coding plan" error payload -- the key is valid and stays enabled; the Quota tab shows an amber "key valid, no active plan" note.
• A dead key returns HTTP 200 with {"code":1000,"msg":"Authentication Failed","success":false} -- detected as an auth rejection and treated like a real 401 (key auto-disabled).

---

Kimi for Coding (Anthropic Format)

Moonshot's Kimi models via a Claude-compatible endpoint.

Seeded models: kimi-for-coding, k3, k3-256k

Model lists change over time -- snapshot from August 2026. Official docs: platform.kimi.ai/docs

Claude Code ready: Because this provider speaks the Anthropic Messages format, you can point Claude Code at it -- use the Claude Code tab in the provider's Prompt for AI dialog.

Quota Windows

Kimi Coding enforces 5-hour, weekly, and monthly quota windows. When a window is full, the API returns HTTP 403 access_terminated_error (not 429). ClawRouter classifies this as quota exhaustion, not an auth error:

• The key enters a timed backoff and is never disabled -- it recovers automatically when the window resets.
• Cooldowns are per-key and window-accurate: every successful quota probe saves a snapshot of that key's windows, and on exhaustion the key cools down until its own exhausted window's reset (5-hour trip > 5-hour reset; weekly trip > weekly reset -- no shared blanket cooldown).
• Monthly billing-cycle exhaustion is detected separately ("Monthly usage cycle exhausted"): the key stays enabled and is retried in ~10-day steps anchored to its last successful use until the cycle renews.
• The provider's Quota tab (?tab=quota, Fetch Quota button) shows one card per enabled key -- sorted usable keys first -- with membership level, region, parallel limit, per-window progress bars (5-hour window, weekly cycle) with remaining quota and reset countdowns, and the booster wallet monthly cap.

---

MiniMax Coding (Anthropic Format)

MiniMax M-series via Claude-compatible endpoints. Two regional presets:

MiniMax Coding (Intl)

MiniMax Coding (China)

Seeded models: MiniMax-M3, MiniMax-M2.7, MiniMax-M2.5, MiniMax-M2.1

Live Model List: MiniMax has a real /v1/models endpoint (on both the international and China hosts) -- Fetch Models retrieves the current catalog, including the -highspeed variants (MiniMax-M2.7-highspeed, MiniMax-M2.5-highspeed, MiniMax-M2.1-highspeed). The seeded list above is used only as a fetch-failure fallback.

Model lists change over time -- snapshot from August 2026. Official docs: platform.minimax.io/docs

---

Alibaba

Four presets covering the coding plans and the general Model Studio API:

Alibaba Coding Plan (CN)

Seeded models: qwen3.7-plus, qwen3.6-plus, qwen3.5-plus, qwen3-max-2026-01-23, qwen3-coder-next, qwen3-coder-plus, glm-5, glm-4.7, kimi-k2.5, MiniMax-M2.5

Alibaba Coding Plan (Intl)

Seeded models: qwen3.7-plus, qwen3.6-plus, qwen3.5-plus, qwen3-max-2026-01-23, qwen3-coder-next, qwen3-coder-plus, glm-5, glm-4.7, kimi-k2.5, MiniMax-M2.5

Alibaba Model Studio (CN)

Seeded models: qwen3.5-plus, qwen3-coder-plus, qwen3-coder-next, qwen-plus, qwen-turbo

Alibaba Model Studio (Intl)

Seeded models: qwen3.5-plus, qwen3-coder-plus, qwen3-coder-next, qwen-plus, qwen-turbo

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list (all four endpoints answer /models). Official docs: help.aliyun.com/en/model-studio

Coding Plan vs Model Studio keys are NOT interchangeable: Coding-Plan keys come from a subscription and only work on the coding{,-intl} endpoints; Model Studio keys are pay-as-you-go and only work on the dashscope{,-intl} endpoints. Mixing them is the usual cause of invalid_api_key.

---

China Providers

Baidu Qianfan

Seeded models: deepseek-v4-pro, deepseek-v4-flash, glm-5.2, kimi-k2.6, qwen3.5-397b-a17b

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list.

Tencent Hunyuan

Seeded models: hunyuan-turbos-latest, hunyuan-t1-latest

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list.

Volcengine Ark

Seeded models: Doubao-Seed-2.0-Code, Doubao-Seed-Code, DeepSeek-V4-Flash, GLM-5.1, Kimi-K2.6

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list. Official docs: volcengine.com/docs/82379

Xiaomi MiMo

Seeded models: mimo-v2.5-pro, mimo-v2.5, mimo-v2-omni, mimo-v2-flash

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list.

Xiaomi MiMo Token Plan

For MiMo subscription keys (tp- prefix).

Seeded models: mimo-v2.5-pro, mimo-v2.5, mimo-v2-pro, mimo-v2-omni

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list.

---

Aggregators & Inference Clouds

SiliconFlow

Seeded models: deepseek-ai/DeepSeek-V4-Pro, deepseek-ai/DeepSeek-R1, Qwen/Qwen3.5-397B-A17B, zai-org/GLM-5.1, moonshotai/Kimi-K2.6

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list. Official docs: docs.siliconflow.cn

Together AI

Seeded models: meta-llama/Llama-3.3-70B-Instruct-Turbo, deepseek-ai/DeepSeek-R1, Qwen/Qwen3-235B-A22B, meta-llama/Llama-4-Maverick-FP8

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list. Official catalog: docs.together.ai/docs/serverless-models

Fireworks AI

Seeded models: accounts/fireworks/models/deepseek-v3p1, accounts/fireworks/models/llama-v3p3-70b-instruct, accounts/fireworks/models/qwen3-235b-a22b

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list. Official docs: docs.fireworks.ai

Featherless

Seeded models: deepseek-ai/DeepSeek-V4-Pro, zai-org/GLM-5.2, moonshotai/Kimi-K2.7-Code

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list.

Hyperbolic

Seeded models: Qwen/QwQ-32B, deepseek-ai/DeepSeek-R1, deepseek-ai/DeepSeek-V3, meta-llama/Llama-3.3-70B-Instruct

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list.

Chutes AI

Nebius AI Studio

Seeded models: meta-llama/Llama-3.3-70B-Instruct

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list.

Venice AI

Seeded models: venice-uncensored-1-2, zai-org-glm-5, qwen3-235b-a22b-instruct-2507, llama-3.3-70b

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list.

Vercel AI Gateway

LLM7

Seeded models: gpt-5.5, deepseek-v4-flash, kimi-k3

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list.

Blackbox AI

Seeded models: claude-sonnet-4.6, gpt-5.4, deepseek-v4-flash, grok-4.3

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list.

Morph

Fast-apply / coding models.

Seeded models: morph-v3-large, morph-v3-fast

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list.

OpenCode Go

OpenCode's paid tier -- same platform as the free OpenCode Zen preset, but with a key.

Seeded models: glm-5.2, kimi-k2.7-code, deepseek-v4-pro, minimax-m3, qwen3.7-max

Model lists change over time -- snapshot from August 2026. Use Models tab > Fetch Models for the live list. Official docs: opencode.ai/docs/zen