All plugins

apinex

APInex as Hermes model providers on one key: `apinex` serves Claude 5.5 and Free-tier Claude models over APInex's native Anthropic Messages API (so Hermes' Anthropic prompt caching and extended thinking apply), and `apinex-openai` serves GPT-6.1, Gemini 4 Argon, DeepSeek V4.1, GLM 5.3, Qwen 3.8, Kimi K3.1, Grok 4.7 and 17 Free-tier ($0.00/1M) chat models over its OpenAI-compatible API. Each /model picker lists that side's tool-capable chat models from the account's GET /v1/models list, omitting non-tool and media endpoints, with 1M/500K context windows and native image input declared per model. Maintained by APInex. Disclosure — registers two model providers at import and nothing else (no tools, hooks, commands or threads). Prompts, tool schemas and APINEX_API_KEY go to https://api.apinex.bond (/v1/messages and /v1/chat/completions) through Hermes' own clients, and APInex forwards each request to the upstream vendor serving the model. The model list is Hermes' standard GET https://api.apinex.bond/v1/models request with the key; the plugin makes no network requests of its own, adds no headers, sends no telemetry, reads or writes no files and stores no credentials.

Apinex-AIModelsv0.1.0Updated 3 hours ago

About

APInex as Hermes model providers on one key: `apinex` serves Claude 5.5 and Free-tier Claude models over APInex's native Anthropic Messages API (so Hermes' Anthropic prompt caching and extended thinking apply), and `apinex-openai` serves GPT-6.1, Gemini 4 Argon, DeepSeek V4.1, GLM 5.3, Qwen 3.8, Kimi K3.1, Grok 4.7 and 17 Free-tier ($0.00/1M) chat models over its OpenAI-compatible API. Each /model picker lists that side's tool-capable chat models from the account's GET /v1/models list, omitting non-tool and media endpoints, with 1M/500K context windows and native image input declared per model. Maintained by APInex. Disclosure — registers two model providers at import and nothing else (no tools, hooks, commands or threads). Prompts, tool schemas and APINEX_API_KEY go to https://api.apinex.bond (/v1/messages and /v1/chat/completions) through Hermes' own clients, and APInex forwards each request to the upstream vendor serving the model. The model list is Hermes' standard GET https://api.apinex.bond/v1/models request with the key; the plugin makes no network requests of its own, adds no headers, sends no telemetry, reads or writes no files and stores no credentials.

hermes-provider-apinex

APInex model providers for Hermes Agent.

APInex is a unified AI API gateway: one sk-apx-... API key for Claude 5.5, GPT-6.1, Gemini 4 Argon, DeepSeek V4.1, Qwen 3.8, GLM 5.3, Kimi K3.1, Grok 4.7, 19 Free-tier chat models ($0.00/1M), and 8 built-in Agent HTTP tools (Web Search, Web Contents, Deep Research, Twitter/X Read API, Solana RPC, Captcha Solver, Translator, Voice STT/TTS). Top up with USDT / crypto or card — no foreign bank card or VPN required.

Hermes speaks one wire protocol per provider, so this plugin registers two providers on the same APINEX_API_KEY:

| Provider | Models | Wire protocol | Endpoint |
|---|---|---|---|
| apinex | Claude (claude-*, free/claude-*) | Native Anthropic Messages API | https://api.apinex.bond/v1/messages |
| apinex-openai | GPT, Gemini, DeepSeek, GLM, Qwen, Kimi, Grok and Free-tier chat models | OpenAI-compatible Chat Completions | https://api.apinex.bond/v1 |

Claude runs over the native Anthropic Messages protocol so Hermes applies Anthropic prompt caching (cache_control breakpoints) and extended thinking (agent.reasoning_effort) exactly as it does for Anthropic directly.

The plugin registers these two model providers and nothing else: no tools, hooks, middleware, commands, background threads or processes. Hermes core files are not touched.

Maintained by APInex.

По-русски. Плагин добавляет APInex в Hermes как два провайдера на одном ключе APINEX_API_KEY: apinex (модели Claude через нативный Anthropic Messages API с поддержкой prompt caching и thinking) и apinex-openai (GPT-6.1, Gemini 4 Argon, DeepSeek V4.1, GLM 5.3, Qwen 3.8, Kimi K3.1, Grok 4.7 и 17 бесплатных моделей free/* за $0.00/1M через OpenAI-совместимый API). Работает из России и СНГ без VPN, пополнение баланса в USDT / крипте без зарубежной банковской карты, при регистрации начисляются бесплатные токены. Документация на русском: apinex.bond/ru/integrations/hermes.

简体中文。 本插件将 APInex 作为两个模型提供商添加到 Hermes，共用同一个 APINEX_API_KEY：apinex 通过原生 Anthropic Messages API（/v1/messages）运行 Claude 5.5 与免费 free/claude-* 模型（原生支持 Prompt Caching 缓存断点与深度思考）；apinex-openai 通过 OpenAI 兼容接口（/v1/chat/completions）运行 GPT-6.1、Gemini 4 Argon、DeepSeek V4.1、Qwen 3.8、GLM 5.3、Kimi K3.1、Grok 4.7 及 17 款 $0.00/1M 免费模型。支持 USDT 充值、免绑海外信用卡、国内直连免翻墙。中文集成指南：apinex.bond/zh/integrations/hermes。

What you get

  • --provider apinex and --provider apinex-openai, both providers in hermes model and /model, hermes setup and hermes doctor checks, all wired automatically by Hermes from the provider profiles.
  • /model pickers built from your account's live model list (GET https://api.apinex.bond/v1/models):
    • 5 Claude models under apinex (claude-sonnet-5.5, claude-opus-5.5, claude-fable-5.1, free/claude-sonnet-4.6, free/claude-opus-4.6).
    • 34 tool-capable chat models under apinex-openai (including 17 free/* models at $0.00/1M).
    • Non-tool models (drex-v2.0, jev-1.13, free/jev-1.13) and image/video/audio/embedding endpoints are filtered out so Hermes never selects a route without tool calling.
  • Context windows (1,000,000 tokens across Claude, GPT, Gemini, DeepSeek, GLM, Qwen, Kimi; 500,000 on grok-4.7) and native image input (supports_vision) declared per model so Hermes does not fall back to generic defaults.
  • Zero-cost auxiliary models for Hermes background tasks (context compression, session titles): free/claude-sonnet-4.6 on apinex and free/deepseek-v4.1-flash on apinex-openai.

Why use the plugin instead of a manual custom_providers entry?

| Feature | Manual custom_providers in config.yaml | hermes-provider-apinex plugin |
|---|---|---|
| Setup | Edit YAML manually, set context_length, list model IDs by hand | hermes plugins install apinex |
| Claude wire protocol | Forced onto /v1/chat/completions (no Anthropic prompt-cache breakpoints) | Native /v1/messages on apinex with automatic prompt caching & thinking |
| /model picker | Unfiltered list (includes non-tool models like drex-v2.0 / jev-1.13) | Filtered to the 39 tool-capable chat models |
| Context windows & vision | Single static context_length override | Per-model 1M / 500K context & supports_vision metadata |
| Background compression model | Uses your paid main model unless configured | Defaults to $0.00/1M free/claude-sonnet-4.6 / free/deepseek-v4.1-flash |

Requirements

  • Hermes Agent 0.21.5 or newer. The apinex provider uses Hermes' Anthropic transport, which needs the anthropic Python package; Hermes installs it on first use (or ships it with its anthropic extra).
  • An APInex API key in APINEX_API_KEY. Create an account at apinex.bond/register (includes free trial tokens, no credit card required) and copy your sk-apx-... key from the dashboard.
  • The plugin itself has no Python dependencies beyond the standard library.

Install

From the Hermes plugin catalog:

hermes plugins install apinex

Or install directly from GitHub into your Hermes home:

git clone https://github.com/Apinex-AI/hermes-provider-apinex.git
mkdir -p ~/.hermes/plugins/model-providers
cp -r hermes-provider-apinex/apinex ~/.hermes/plugins/model-providers/apinex

Use

Set APINEX_API_KEY in your shell or in ~/.hermes/.env:

export APINEX_API_KEY="sk-apx-..."

# Native Anthropic Messages API (Claude 5.5 + free/claude-*)
hermes --provider apinex --model claude-sonnet-5.5

# OpenAI-compatible API (GPT-6.1, Gemini 4 Argon, DeepSeek V4.1, Free tier)
hermes --provider apinex-openai --model free/deepseek-v4.1-flash
hermes --provider apinex-openai --model gpt-6.1-sol

Or run hermes model, select APInex or APInex (OpenAI-compatible), and pick a model interactively.

In ~/.hermes/config.yaml:

model:
  provider: apinex
  default: claude-sonnet-5.5

Or for the OpenAI-compatible side:

model:
  provider: apinex-openai
  default: free/deepseek-v4.1-flash

Supported models & pricing snapshot

Full live catalog and benchmarks: apinex.bond/models and apinex.bond/pricing.

apinex — Native Anthropic Messages API (https://api.apinex.bond/v1/messages)

| Model ID | Context | Vision | Tool Calling | Input / 1M | Output / 1M |
|---|---|---|---|---|---|
| claude-sonnet-5.5 | 1,000,000 | Yes | Yes | $0.40 | $0.40 |
| claude-opus-5.5 | 1,000,000 | Yes | Yes | $0.50 | $0.50 |
| claude-fable-5.1 | 1,000,000 | Yes | Yes | $2.00 | $2.00 |
| free/claude-sonnet-4.6 | 1,000,000 | Yes | Yes | $0.00 | $0.00 |
| free/claude-opus-4.6 | 1,000,000 | Yes | Yes | $0.00 | $0.00 |

apinex-openai — OpenAI-compatible API (https://api.apinex.bond/v1)

| Model ID | Family | Context | Vision | Input / 1M | Output / 1M |
|---|---|---|---|---|---|
| free/deepseek-v4.1-flash | DeepSeek | 1,000,000 | Yes | $0.00 | $0.00 |
| free/glm-5.3-flash | GLM | 1,000,000 | Yes | $0.00 | $0.00 |
| free/qwen-3.8-max | Qwen | 1,000,000 | Yes | $0.00 | $0.00 |
| free/gemini-3.8-flash | Gemini | 1,000,000 | Yes | $0.00 | $0.00 |
| free/gemini-3.1-pro | Gemini | 1,000,000 | Yes | $0.00 | $0.00 |
| free/gpt-6-luna | GPT | 1,000,000 | Yes | $0.00 | $0.00 |
| free/kimi-k3 | Kimi | 1,000,000 | Yes | $0.00 | $0.00 |
| free/minimax-m3.1-flash | MiniMax | 1,000,000 | Yes | $0.00 | $0.00 |
| gpt-6.1-sol | GPT | 1,000,000 | Yes | $0.30 | $0.30 |
| gpt-6.1-terra | GPT | 1,000,000 | Yes | $0.17 | $0.17 |
| gpt-6-astra | GPT | 1,000,000 | Yes | $0.50 | $0.50 |
| gpt-6-luna | GPT | 1,000,000 | Yes | $0.075 | $0.075 |
| gemini-4-argon | Gemini | 1,000,000 | Yes | $0.60 | $0.60 |
| gemini-3.1-pro | Gemini | 1,000,000 | Yes | $0.15 | $0.15 |
| gemini-3.8-flash | Gemini | 1,000,000 | Yes | $0.10 | $0.10 |
| deepseek-v4.1-flash | DeepSeek | 1,000,000 | Yes | $0.06 | $0.06 |
| deepseek-v4-pro | DeepSeek | 1,000,000 | No | $0.07 | $0.07 |
| deepseek-v4-flash | DeepSeek | 1,000,000 | No | $0.05 | $0.05 |
| glm-5.3-flash | GLM | 1,000,000 | Yes | $0.05 | $0.05 |
| glm-5.3 | GLM | 1,000,000 | No | $0.15 | $0.15 |
| kimi-k3.1-pro | Kimi | 1,000,000 | Yes | $0.30 | $0.40 |
| grok-4.7 | Grok | 500,000 | Yes | $0.30 | $0.30 |
| muse-spark-1.3 | Muse | 1,000,000 | Yes | $0.30 | $0.30 |
| hy4-preview | Hunyuan | 1,000,000 | Yes | $0.30 | $0.30 |

Plus 9 additional free/* routes (free/deepseek-v4-flash-0731, free/deepseek-v4-pro-0813, free/glm-5.3, free/glm-5.3-flashx, free/hy4, free/kimi-k2.8, free/mimo-v2.6-flash, free/mimo-v2.6-pro, free/minimax-m3, free/muse-spark-1.3).

Companion APInex skills & ecosystem for Hermes

Your APINEX_API_KEY also unlocks 8 HTTP agent tools on https://api.apinex.bond/v1/tools/*. You can discover and install the official APInex workflow skills directly in Hermes via well-known:

hermes skills search https://apinex.bond --source well-known

Install any skill with one command:

hermes skills install well-known:https://apinex.bond/.well-known/skills/apinex-quickstart
hermes skills install well-known:https://apinex.bond/.well-known/skills/apinex-cli
hermes skills install well-known:https://apinex.bond/.well-known/skills/apinex-web-search
hermes skills install well-known:https://apinex.bond/.well-known/skills/apinex-twitter-api
hermes skills install well-known:https://apinex.bond/.well-known/skills/apinex-solana-captcha-voice
hermes skills install well-known:https://apinex.bond/.well-known/skills/apinexgate-reseller

Security, privacy and footprint

  • Two provider profiles only. Importing apinex calls register_provider() twice (apinex and apinex-openai). It registers no tools, hooks, middleware, slash commands, CLI subcommands, background threads or child processes.
  • Network. The plugin makes no network requests of its own. Model listing uses Hermes' ProviderProfile.fetch_models(), which sends GET https://api.apinex.bond/v1/models with Authorization: Bearer $APINEX_API_KEY. Chat requests are sent by Hermes' own Anthropic and OpenAI clients to https://api.apinex.bond/v1/messages (for apinex) and https://api.apinex.bond/v1/chat/completions (for apinex-openai), which forward each request to the upstream model provider. If you set model.base_url to another host, requests go to that host instead.
  • Credentials and files. Hermes reads APINEX_API_KEY from your environment or ~/.hermes/.env and passes it to its clients. The plugin never reads, writes, logs, caches or stores the key or any local files.
  • No extra headers or telemetry. default_headers is empty ({}). No analytics, telemetry or crash reports are sent.
  • Terms & privacy: apinex.bond/privacy and apinex.bond/terms.

Tests

Offline unit and packaging tests (standard library only, no Hermes checkout or network required):

python3 -m unittest discover -s tests -v

Changelog

0.1.0

  • Initial release: apinex (native Anthropic Messages API) and apinex-openai (OpenAI-compatible Chat Completions) providers on a single APINEX_API_KEY, with per-model context windows and vision capabilities for all 39 tool-capable APInex chat models.

README from the reviewed commit d838363; it updates when the author re-pins.