wiroaiWiro AI for Hermes
Use the Wiro AI LLM gateway as a model provider in
Hermes Agent. One OpenAI-compatible API and one key
reach GPT, Claude, Gemini, DeepSeek, GLM, Kimi and Grok models. Requires Hermes 0.22.6 or later
and a Wiro AI API key. The plugin has no dependencies.
Connect
hermes plugins install wiro
hermes config set WIRO_API_KEY <your-key>
hermes model
Before the catalog listing lands, install from the repository instead:
hermes plugins install "https://github.com/wiroai/Wiro-Hermes#wiro".
Create the key under your project at wiro.ai/panel/project. For a
Signature project, the value is YOUR_API_KEY:YOUR_API_SECRET. In hermes model, select
Wiro AI and pick a model. One-off use:
hermes chat --provider wiro -m claude/sonnet-5
Models
hermes model lists the live catalog, GET https://llm.wiro.ai/v1/models. The models below are
its fallback list, with values from that catalog. Each one calls tools and takes a reasoning level.
| Model | Context window | Image input |
|---|---:|---|
| openai/gpt-5-4-mini (default) | 400,000 | Yes |
| claude/sonnet-5 | 1,000,000 | Yes |
| google/gemini-3-flash | 1,048,576 | Yes |
| deepseek/v4-flash | 1,048,576 | No |
| glm/5-3 | 1,000,000 | Yes |
| moonshot/kimi-k3-1-pro | 1,048,576 | Yes |
| xai/grok-4-5 | 500,000 | Yes |
| openai/gpt-6-sol | 1,050,000 | Yes |
| openai/gpt-5-4 | 1,050,000 | Yes |
| claude/opus-5-5 | 1,000,000 | Yes |
| google/gemini-3-pro | 1,048,576 | Yes |
| deepseek/v4-pro | 1,048,576 | No |
| openai/gpt-6-luna (background tasks) | 1,050,000 | Yes |
| openai/gpt-5-4-nano | 400,000 | Yes |
openai/gpt-6-luna also describes images for Hermes when the chat model reads text only. Any other
model in the catalog works too, with its own default reasoning. Prices are on each model's page
in the Wiro AI catalog and in the pricing field
of GET /v1/models. Usage is paid as you go from your Wiro wallet.
Reasoning
Hermes' reasoning level (/reasoning, --reasoning or agent.reasoning_effort) is sent as
reasoning_effort for the models listed above. With no level set nothing is sent, so the model's
default applies. A level a model does not offer is moved to the nearest one it does, and the
response names the change in its x-wiro-adjusted-parameters header. Hermes' ultra is sent as
max.
When reasoning is off, the OpenAI models and Kimi K3.1 Pro get none. For the other listed models
none is not one of their levels, so they get their lowest level instead: low for Claude,
Gemini 3 Pro and Grok 4.5, minimal for Gemini 3 Flash, DeepSeek and GLM.
Configuration
| Variable | Purpose |
|---|---|
| WIRO_API_KEY | API key (required) |
The provider id is wiro, with the aliases wiro-ai and wiroai.
Scope
Prompts, tool definitions and results, attached images and model-list requests go to
https://llm.wiro.ai/v1 through Hermes' own client, with the key Hermes reads. Usage is paid from
your Wiro wallet. The plugin registers one model provider and nothing else: no tools, hooks,
background work, telemetry, file access or self-update. Update it with hermes plugins update wiro.
MIT license.
README from the reviewed commit ac9cd5f; it updates when the author re-pins.
