morph-hermes
A Hermes Agent model-provider
plugin for Morph. Morph serves open models such as Kimi,
GLM and DeepSeek on an OpenAI-compatible API.
Install
hermes plugins install morphllm
Setup
-
Create an API key at
morphllm.com/dashboard/api-keys. -
Save it, either in
~/.hermes/.envor the process environment:MORPH_API_KEY=sk-...Optional:
MORPH_BASE_URLpoints at another Morph endpoint. Default is
https://api.morphllm.com/v1. -
Select Morph in Hermes:
hermes model→ Morph (provider:morph).
The model picker lists the chat models your key can call, live from
/v1/models. Morph's Fast Apply, WarpGrep and compaction models are left out
because they are not chat models.
Reasoning
When you set a reasoning effort in Hermes, it is sent as reasoning_effort.
Turning reasoning off sends reasoning_effort: "none". With nothing set, each
model runs at its served default.
What it does on your machine
- Makes HTTPS requests only to
api.morphllm.com(or yourMORPH_BASE_URL):
/v1/modelsfor the model list and the normal chat requests. - Reads
MORPH_API_KEYthrough Hermes's provider credentials. It stores
nothing and reads no other files. - No telemetry, no shell commands, no background processes, no self-updating
code.
README from the reviewed commit a5ca347; it updates when the author re-pins.

morphllm