Obscurify.ai provider for Hermes Agent

Adds Obscurify.ai to Hermes Agent as a model provider: pick Obscurify.ai in hermes model, or pass --provider obscurify.

Obscurify.ai is a privacy-focused, OpenAI-compatible API to 200+ models (DeepSeek, GLM, Kimi, Qwen, GPT, Claude, Gemini, Llama and more). Text requests only use zero-data-retention routes, prompts and responses aren't stored on its servers, and accounts are a random account number with no email.

What the plugin adds over a custom endpoint:

• A model list Hermes can run. It comes from the public https://obscurify.ai/v1/models catalog, minus decision models (they only answer structured questions) and models under the 64K-token context Hermes needs.
• Context windows from the same catalog.
• Reasoning effort sent as reasoning_effort (none through max). Models that don't reason ignore it.

Install

hermes plugins install obscurify-provider

Or straight from this repository, pinned to a reviewed commit:

hermes plugins install https://github.com/Obscurify-ai/hermes-obscurify-provider --ref FULL_COMMIT_SHA

Needs Hermes 0.22.5 or newer and no other Python packages.

Configure

1. Create an account at obscurify.ai. You get a random account number: no email, no password.
2. Add a balance. The API works on every plan with a positive balance. Frontier models need the Casual plan or higher, and uncensored models need Pro.
3. On obscurify.ai/account, click Show API Key, then put the key in your Hermes profile's environment:

export OBSCURIFY_API_KEY="sk-obs1_..."

Pick the provider and a model:

hermes model --provider obscurify

or run a one-shot request:

hermes -z "Reply with exactly: ok" --provider obscurify -m deepseek/deepseek-v4-flash

Good starting models: deepseek/deepseek-v4-flash (default auxiliary model), z-ai/glm-5.3, moonshotai/kimi-k2-0905, qwen/qwen3-coder. Prices per model: obscurify.ai/model_cards.

Disclosure

The plugin registers one model provider at import and nothing else: no tools, hooks, commands or background threads. Chat requests go to https://obscurify.ai/v1 through Hermes' own client, with OBSCURIFY_API_KEY, which Hermes reads. The plugin itself only fetches the public model catalog (GET https://obscurify.ai/v1/models, no credentials), at most once an hour per process, to build the model list and context windows. It reads and writes no files and starts no processes. Using a model costs money from your Obscurify.ai balance at that model's price.

Tested

With Hermes 0.22.5 and main (2026-10-28), using a Pay As You Go key:

• hermes plugins validate --install-deps . passes, including the security scan.
• hermes -z ... --provider obscurify -m meta-llama/llama-3.3-70b-instruct answers.
• With deepseek/deepseek-v4-flash, Hermes wrote a file through its file tool.

License

MIT