Run Hermes on Mingles Router
A step-by-step wizard for pointing the Hermes Agent at Mingles Router — whether you are installing it from scratch or already have it running.
Pick your model and paste your key — every command and value on this page updates live.
Endpoint: https://router.mingles.ai/v1 · key stays in your browser ·
no key entered? commands show <your-key>.
Install + interactive setup
1. Install Hermes
One command. Works on macOS and Linux.
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash2. Start the setup
Run hermes (or hermes setup). At the first prompt choose Full setup — bring your own keys, not Quick Setup. In the provider list, scroll to the bottom and pick Custom endpoint (enter URL manually).
hermes3. Answer the four prompts
Hermes asks for the base URL, key, compatibility mode and model — use the exact values in the table below. The endpoint is verified against /v1/models as you type the URL; you should see “3 model(s) visible”.
4. Verify & chat
hermes doctor checks the connection; then just run hermes to start a conversation.
hermes doctor
hermesExact values Hermes asks for
| API base URL | https://router.mingles.ai/v1 |
| API key | <your-key> |
| API compatibility mode | 2 — Chat Completions |
| Model | moonshotai/Kimi-K2.6 |
| Context length | 200000 |
Point an existing Hermes at Mingles Router — no wizard
Set the four values with hermes config set
Hermes stores model config in ~/.hermes/config.yaml and secrets in ~/.hermes/.env. The custom provider speaks the OpenAI Chat Completions dialect.
hermes config set model.provider custom
hermes config set model.base_url https://router.mingles.ai/v1
hermes config set model.api_key <your-key>
hermes config set model.default moonshotai/Kimi-K2.6
hermes config set model.context_length 200000model.api_key is scoped to this endpoint. Prefer a global key? Put OPENAI_API_KEY=… in ~/.hermes/.env instead — the custom provider falls back to it.
Or edit ~/.hermes/config.yaml directly
Same result, hand-written:
model:
provider: custom
base_url: "https://router.mingles.ai/v1"
api_key: "<your-key>"
default: "moonshotai/Kimi-K2.6"
context_length: 200000Switch model
All three models share the same endpoint, so switching is just a model id — no new key, no re-setup. Inside a chat use /model, or persist it from the shell:
/model custom:moonshotai/Kimi-K2.6hermes config set model.default moonshotai/Kimi-K2.6Before you file a bug: read the limits
Reasoning models spend the output budget on internal thinking, output is capped at 8192 tokens, there is no KV cache, and there are no built-in web tools. Most “it broke on Mingles Router” reports are one of these. See Model limits & behavior →
Frequently asked
Which API compatibility mode do I pick? +
Chat Completions (option 2). Mingles Router speaks the OpenAI /v1/chat/completions dialect. Do not pick Responses/Codex or Anthropic Messages.
What provider name does Hermes use for a custom endpoint? +
custom. In ~/.hermes/config.yaml that is model.provider: "custom" with model.base_url pointing at the gateway. Aliases ollama / vllm / llamacpp also map to custom.
What context length should I enter? +
Set model.context_length to 200000 for Kimi-K2.6. Hermes auto-detects context length from a provider catalog, but it does not know our custom endpoint, so set it manually. Output is separately capped at 8192 tokens.
Where is the API key stored? +
Either in ~/.hermes/config.yaml (model.api_key) or in ~/.hermes/.env as OPENAI_API_KEY — the custom provider falls back to OPENAI_API_KEY. The key never leaves your machine.
How do I switch between Kimi, Qwen and MiniMax? +
They all live on the same endpoint, so just change the model id: /model custom:<model-id> in a session, or hermes config set model.default <model-id>. No new keys needed.