How to Use Agnes’s Free Multimodal Model Across All Major Agent Platforms

Cost Context

Claude subscription costs about $200 USD per month (iOS adds $50) and Codex about $100 USD per month (iOS adds $25), totaling over $2000 USD annually. Token consumption in agent workflows grows rapidly, as discussed in the author’s earlier “Token Economics” article.

Agnes Model Suite

Agnes provides three multimodal models (text, image, video) that became free of charge on 1 June. Recent leaderboard placements:

Text model agnes-2.0-flash ranked mid‑range on Claw‑Eval, behind Claude Opus 4.6 and Sonnet 4.6, but outperformed Gemini 3 Flash and MiniMax M2.7 on the Pass³ metric.

Image model placed in the middle of the Artificial Analysis blind‑review ranking.

Video model appeared on the same provider’s video‑generation list.

The provider claims a global AI‑lab ranking of 9th.

Pre‑Free Pricing

Text: $0.03 per 1 M input tokens, $0.15 per 1 M output tokens.

Image: $3 per 1 k images (≈10× cheaper than typical $30‑level competitors).

Video: $0.30 per minute (≈$0.03 for a 10‑second 720p clip).

Gateway and Authentication

All tools use the same OpenAI‑compatible endpoint: https://apihub.agnes-ai.com/v1 Create an API key at https://platform.agnes-ai.com/. The eight supported tools fall into two integration categories:

OpenAI‑format tools (OpenClaw, HermesAgents, WorkBuddy, Cherry Studio, Opencode, Codex++) – configure gateway URL, API key, and model name agnes-2.0-flash.

Anthropic‑protocol tools (Claude Code, Claude Desktop) – require protocol conversion via cc-switch (v3.16.1).

1. OpenClaw

Run openclaw config, select Local → Model → Custom Provider , then set:

API Base URL: https://apihub.agnes-ai.com/v1
API Key: YOUR_KEY
Model: agnes-2.0-flash

Save and test with a simple conversation.

2. HermesAgents

hermes config set model.provider custom
hermes config set model.base_url https://apihub.agnes-ai.com/v1
hermes config set model.api_key YOUR_API_KEY
hermes config set model.default agnes-2.0-flash

Ask a question to verify connectivity.

3. Claude Code (via cc‑switch)

Download cc-switch from https://github.com/farion1231/cc-switch. In the Claude CLI tab, add a custom Claude supplier with:

API Key: YOUR_KEY

Base URL: https://apihub.agnes-ai.com/v1

API format: OpenAI Chat Completions

Auth field: default ANTHROPIC_AUTH_TOKEN

Map Sonnet, Opus, Haiku to agnes-2.0-flash and add the following parameter filter to avoid incompatibilities:

{
  "allowed_openai_params": ["thinking", "context_management"],
  "litellm_settings": {"drop_params": true}
}

Enable the Claude route in the local router and test with a dialogue.

4. Claude Desktop (via cc‑switch)

Install the appropriate binary (Windows .exe, macOS .dmg, Linux AppImage). In cc‑switch’s Claude Desktop tab, add a custom provider with the same API key and base URL, turn on the model‑mapping switch, and map Sonnet, Opus, Haiku to agnes-2.0-flash. Save and verify with a conversation.

5. WorkBuddy

In WorkBuddy v4.24.5, click Auto → Custom Model → Other (custom) and enter:

API Base URL: https://apihub.agnes-ai.com/v1
API Key: YOUR_KEY
Model: agnes-2.0-flash

Test the text model, then create Skills for image and video generation using commands such as agnes-image-gen skill and agnes-video-gen skill.

6. Cherry Studio

Add a new OpenAI provider named “Agnes”, set the API key and base URL, fetch the model list, select agnes-2.0-flash, and save. Use the model in a new conversation; for image/video, create Skills and bind them to the agent.

7. Opencode

Edit ~/.config/opencode/opencode.jsonc to include:

{
  "$schema": "https://opencode.ai/config.json",
  "model": "agnes/agnes-2.0-flash",
  "provider": {
    "agnes": {
      "options": {
        "apiKey": YOUR_KEY,
        "baseURL": "https://apihub.agnes-ai.com/v1"
      },
      "models": {
        "agnes-2.0-flash": {"id": "agnes-2.0-flash"}
      }
    }
  }
}

Restart Opencode and test with a simple query.

8. Codex++

Clone the repository https://github.com/BigPizzaV3/CodexPlusPlus, install the package, and on macOS remove quarantine flags:

sudo xattr -rd com.apple.quarantine "/Applications/Codex++.app"
sudo xattr -rd com.apple.quarantine "/Applications/Codex++ 管理工具.app"

In the Codex++ manager, add a provider named “Agnes” with Access Mode “API Only”, model “agnes-2.0-flash”, Base URL https://apihub.agnes-ai.com/v1, and your API key. Enable the provider, start Codex++, and verify a response.

Summary

All eight platforms share a two‑step integration: supply the gateway URL and API key. Only Claude‑based tools need the cc‑switch conversion layer. Agnes’s free‑forever policy reduces inference cost to roughly one‑tenth of typical industry rates, making advanced multimodal models accessible to budget‑constrained users.