Hermes Agent custom provider: an uncensored model in config.yaml

Point Hermes Agent at an uncensored model: one providers entry in ~/.hermes/config.yaml, the key in .env, /model to switch, the cost of a turn and the errors.

Steps checked November 6, 2026 · prices as of November 16, 2026

Any OpenAI-compatible endpoint can be a named entry under providers in Hermes Agent’s ~/.hermes/config.yaml. Add one, and the agent reads files and runs shell commands on an uncensored model, paid per token from a prepaid balance. OpenClaw and OpenCode take the same ids.

What you need

  • Hermes Agent installed.
  • A key from the console. Give Hermes a key of its own, so its traffic is easy to pick out on the usage page.
  • A model id from the models page; the YAML below opens with a short list of them.

Three steps to a first reply

  1. Make the key in the console and write it to the env file as a single line:
.env
1# ~/.hermes/.env
2UNRESTRICTABLE_API_KEY=sk-unr-live-…
  1. Put the provider in config.yaml. The ids listed under models are only a first set; every id on the models page fits the same mapping:
config.yaml
1# ~/.hermes/config.yaml
2model:
3 default: glm-5.3-uncensored
4 provider: custom:unrestrictable
5
6providers:
7 unrestrictable:
8 api: https://api.unrestrictable.ai/v1
9 key_env: UNRESTRICTABLE_API_KEY
10 transport: chat_completions
11 default_model: glm-5.3-uncensored
12 models:
13 glm-5.3-uncensored: {}
14 glm-5.3-flash-uncensored: {}
15 deepseek-v4.1-flash-abliterated: {}
16 deepseek-v4-pro-uncensored: {}
17 glm-5.3-abliterated: {}
18 kimi-k3.1-pro-uncensored: {}
  1. Start hermes and type something. Within seconds the console’s usage page lists the request with its token counts and its charge.

Prefer the shell? These commands write the same block. A dotted path lands in config.yaml; an UPPER_SNAKE name lands in .env:

shell
1hermes config set UNRESTRICTABLE_API_KEY sk-unr-live-…
2hermes config set providers.unrestrictable.api https://api.unrestrictable.ai/v1
3hermes config set providers.unrestrictable.key_env UNRESTRICTABLE_API_KEY
4hermes config set providers.unrestrictable.transport chat_completions
5hermes config set model.provider custom:unrestrictable
6hermes config set model.default glm-5.3-uncensored

To change model mid-session, use /model. Without --global the choice ends with the session; with it, Hermes writes the choice into config.yaml:

hermes
1/model custom:unrestrictable:glm-5.3-flash-uncensored
2/model custom:unrestrictable:deepseek-v4.1-flash-abliterated --global

To talk to the same models with nothing installed, open /chat in the browser.

Where everything lives

Two files under ~/.hermes/: config.yaml for settings and .env for secrets. Precedence runs CLI flags, then config.yaml, then .env, then the built-in defaults. The key_env field of a providers entry names the variable that holds the key, which keeps the key itself out of config.yaml. You may still meet a custom_providers list in older write-ups; Hermes reads it and hermes update converts it, but the mapping on this page is the current form.

Each key, explained

api
https://api.unrestrictable.ai/v1
the endpoint including /v1; base_url works as an alias
key_env
UNRESTRICTABLE_API_KEY
the .env variable holding the full sk-unr- key
transport
chat_completions
the OpenAI chat format; every build on the models page speaks it
default_model
glm-5.3-uncensored
the id the entry opens with
models
id: {}
settings per model, keyed by id; a context_length under an id replaces the detected value
model.provider
custom:unrestrictable
picks the entry; model.default names the starting id

Choosing the model

Each file read, shell command and subagent in Hermes is a tool call, so on an unfamiliar id give it one small task first and watch that a tool actually runs. Then look at input price: every turn resends the whole conversation plus all tool output, so input is what a session costs. The table prices a single turn, 30,000 tokens in and 1,000 out, using the catalog on 2026-11-16.

ModelPer turnPer 30 turns
GLM-5.3 Uncensored$0.0093$0.28
GLM-5.3 Flash Uncensored$0.0035$0.11
DeepSeek V4.1 Flash Abliterated$0.03$0.81
DeepSeek V4 Pro Uncensored$0.03$0.93
GLM-5.3 Abliterated$0.07$2.10
Kimi K3.1 Pro Uncensored$0.04$1.22
Uncensored models and their prices per million tokens
GLM-5.3 Uncensoredglm-5.3-uncensoredUncensored$0.282$0.846
GLM-5.3 Flash Uncensoredglm-5.3-flash-uncensoredUncensored$0.1053$0.3685
DeepSeek V4.1 Flash Abliterateddeepseek-v4.1-flash-abliteratedAbliterated$0.8488$1.6975
DeepSeek V4 Pro Uncensoreddeepseek-v4-pro-uncensoredUncensored$0.97$1.94
GLM-5.3 Abliteratedglm-5.3-abliteratedAbliterated$2.1825$4.365
Kimi K3.1 Pro Uncensoredkimi-k3.1-pro-uncensoredUncensored$1.164$5.82

Prices per million tokens, read from the live catalog on 2026-11-16.

If the work is code, the uncensored coding AI guide sets the code builds side by side and prices a whole day on each.

Test it

Run hermes status: it prints the model, the provider in use and which providers hold a credential, never the key itself. Then, in the same shell, ask the API for the model list with that key:

shell
1hermes status
2curl https://api.unrestrictable.ai/v1/models -H "Authorization: Bearer $UNRESTRICTABLE_API_KEY"

Last, one request in Hermes’ own shape. The -i flag shows the headers: x-uplink-request-id identifies the request, and on a non-streamed reply x-uplink-charge states the cost in USD.

shell
1curl -i https://api.unrestrictable.ai/v1/chat/completions \
2 -H "Authorization: Bearer $UNRESTRICTABLE_API_KEY" \
3 -H "Content-Type: application/json" \
4 -d '{"model": "glm-5.3-uncensored", "max_tokens": 16, "messages": [{"role": "user", "content": "Reply with the single word ok."}]}'

When it fails

  • 401 invalid_api_key: the key is mistyped, revoked, or never reaches the request. Look in ~/.hermes/.env for the line, make sure key_env points at that variable, and paste the key from the console again.
  • 402 insufficient_balance: not enough balance for the call; add credit in the console. 402 monthly_limit_reached: the key has reached the monthly limit you gave it.
  • 404 model_not_found: model.default holds an id the models page does not list. Copy one from the page and apply it through hermes config set model.default <id>.
  • 429 rate_limited: your key’s own rate limit; retry-after in the response says when to try again.
  • A 400 on the very first run: the Hermes FAQ blames a mismatched model name. Run hermes config show, then hold model.default against the list on the models page.
  • No custom endpoint is read: a leftover custom_providers key in config.yaml, which is no longer a YAML list. Move it into the providers mapping shown here and run hermes doctor.

Where the facts come from

Field names are those of the providers page and the configuration guide in the Hermes Agent docs, both read 2026-11-06. Every price on this page is the live catalog’s at render time.

Questions

It does. A named entry under providers in ~/.hermes/config.yaml, with api, key_env and transport chat_completions, is all it takes. The YAML on this page aims that entry at https://api.unrestrictable.ai/v1, which puts every build on the models page within reach.

Either edit model.default (the id) and model.provider (custom:unrestrictable), or type /model custom:unrestrictable:<id> during a session; --global makes that stick in config.yaml. Paste the id from the models page instead of typing it.

On one line of ~/.hermes/.env: UNRESTRICTABLE_API_KEY=... An UPPER_SNAKE name given to hermes config set goes there as well. key_env in the providers entry points at it, so config.yaml holds no key.

Tokens times the model’s price, drawn from a prepaid balance; no subscription is needed. The table on this page prices an agent-sized turn on each build at today’s rate, and a repeated prefix is charged as cached input where the model has a cached rate.

Every model above is on the uncensored model list with its live price per million tokens. Per-token pricing sets the whole catalog side by side, and the family directory groups the builds by base model.

Get a keyPrepaid from $10, no subscription.

New models and price changes, once a month.

New uncensored builds with their prices, price moves, and what the daily model check found.

Join on Telegram