Any API that speaks the OpenAI chat format can become a custom endpoint in LibreChat, declared in librechat.yaml with four things: a name, a key, a base URL and the models to offer. Add one such entry and the uncensored builds appear in LibreChat’s endpoint selector beside the others, paid per token. The ids below also drop into Open WebUI and OpenCode.
What you need
A running LibreChat (the Docker setup is assumed below), a key from the console, and write access to the project root. LibreChat loads librechat.yaml from there, and each custom endpoint is a single list item under endpoints.custom. The key itself belongs in .env; the YAML points at it by variable name, which is the form LibreChat’s docs recommend for apiKey. With Docker, the YAML reaches the container through a bind mount declared in docker-compose.override.yml, and a restart picks it up.
First reply
- Create a key in the console and add it to
.env, the file that sits besidedocker-compose.yml:
# .envUNRESTRICTABLE_API_KEY=sk-unr-live-…- Declare the endpoint in
librechat.yaml. Treat the ids undermodels.defaultas a first pick; every id on the models page is written the same way:
# librechat.yamlversion: 1.3.18cache: trueendpoints: custom: - name: 'Unrestrictable' apiKey: '${UNRESTRICTABLE_API_KEY}' baseURL: 'https://api.unrestrictable.ai/v1' models: default: - 'glm-5.3-uncensored' - 'deepseek-v4-pro-uncensored' - 'deepseek-v4.1-flash-abliterated' - 'glm-5.3-abliterated' - 'glm-5.3-flash-uncensored' fetch: false titleConvo: true titleModel: 'glm-5.3-flash-uncensored' modelDisplayLabel: 'Unrestrictable'- Make sure the override file still mounts the YAML, then restart:
# docker-compose.override.ymlservices: api: volumes: - type: bind source: ./librechat.yaml target: /app/librechat.yamlRun docker compose down && docker compose up -d and the file is read again. Your endpoint shows in the selector under its name, offering the models from models.default.
To try the same models before touching any YAML, the browser chat runs the same catalog with three free turns a day and no account.
Each field, explained
- name
- Unrestrictablewhat the endpoint selector shows; must be present and unique
- apiKey
- ${UNRESTRICTABLE_API_KEY}pulls the sk-unr- key out of .env; user_provided instead asks every user for their own
- baseURL
- https://api.unrestrictable.ai/v1keep the /v1; LibreChat adds /chat/completions on its own, so leave directEndpoint out
- models.default
- the ids abovewhat the model picker offers while fetch is false; required
- models.fetch
- falseset to true, it would fill the picker with every model the key can reach
- titleConvo, titleModel
- glm-5.3-flash-uncensoredturns on automatic chat titles and names the small model that writes them
- modelDisplayLabel
- Unrestrictablethe text next to the AI icon in a chat; defaults to "AI"
Which model
On every turn LibreChat resends the entire conversation in the OpenAI format, and all the builds on the models page accept it. The first cost column prices one turn of 2,000 tokens in and 500 out; the second prices one title request. Both use the catalog prices of 2026-11-16.
| Model | One turn | One title |
|---|---|---|
| GLM-5.3 Uncensored | $0.00099 | $0.0003 |
| DeepSeek V4 Pro Uncensored | $0.0029 | $0.001 |
| DeepSeek V4.1 Flash Abliterated | $0.0025 | $0.00088 |
| GLM-5.3 Abliterated | $0.0065 | $0.0023 |
| GLM-5.3 Flash Uncensored | $0.00039 | $0.00011 |
| GLM-5.3 Uncensored | Uncensored | $0.282 | $0.846 |
| DeepSeek V4 Pro Uncensored | Uncensored | $0.97 | $1.94 |
| DeepSeek V4.1 Flash Abliterated | Abliterated | $0.8488 | $1.6975 |
| GLM-5.3 Abliterated | Abliterated | $2.1825 | $4.365 |
| GLM-5.3 Flash Uncensored | Uncensored | $0.1053 | $0.3685 |
Prove it works
On the host that runs LibreChat, ask the API for the models the key can reach. Each id in models.default must appear:
curl https://api.unrestrictable.ai/v1/models -H "Authorization: Bearer sk-unr-live-…"Then send a single streamed request, as LibreChat would. With -i the headers print, and x-uplink-request-id is the id to quote to support.
curl -i https://api.unrestrictable.ai/v1/chat/completions \ -H "Authorization: Bearer sk-unr-live-…" \ -H "Content-Type: application/json" \ -d '{"model": "glm-5.3-uncensored", "messages": [{"role": "user", "content": "Reply with the single word ok."}], "stream": true}'When both commands succeed and LibreChat still fails, the fault is in the YAML or the mount. Read the container log right after docker compose up.
When it fails
- No endpoint in the selector: either the file was never mounted or the YAML failed to parse. The container log after
docker compose uppoints at the line. 401 invalid_api_key: a wrong or revoked key, or the variable named inapiKeyis not the one spelled in.env.402 insufficient_balance: not enough balance for this request; top up in the console.402 monthly_limit_reached: the key has reached the monthly spending cap you gave it.404 model_not_found: one of the ids inmodels.defaultdoes not exist on the models page, or a display name was written in place of an id.429 rate_limited: the key’s rate limit was hit. All users of a single endpoint share that key and that limit.retry-afterin the response says how long to wait.- The picker lists every model the key can reach:
fetchis true. Make itfalseand curate the list inmodels.default.
Sources
The field names, the environment-variable form of apiKey, the baseURL convention and the mount come from LibreChat’s documentation, read 2026-11-06: Custom Endpoint Object Structure and Custom Endpoints quick start. Every price on this page is rendered from the live catalog.
Questions
librechat.yaml in the project root. The endpoint is one list item under endpoints.custom carrying name, apiKey, baseURL and models. Under Docker, docker-compose.override.yml mounts that file into the container, and a restart loads it.
It should: https://api.unrestrictable.ai/v1. LibreChat adds /chat/completions after it, exactly as its own OpenAI-compatible examples do with a /v1 base. directEndpoint exists for a URL that already is the full completions path.
They can. Put user_provided in apiKey and each user types a key into the endpoint's settings, drawing on their own balance. With the env-var reference, everyone on the instance chats on the single key you configured.
All of them. One key covers the whole models page, and each build there takes an OpenAI-format call to /v1/chat/completions, so models.default accepts any id from that page, spelled the same way. See the models page.
LibreChat is free software, so the only cost is tokens, at the live prices in the table, drawn from a prepaid balance with no subscription required. Pointing titleModel at a small build keeps the title requests cheap.
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.