Cline’s OpenAI Compatible provider wants three things from this API: a base URL, a key and a model id. Its Model Configuration panel then wants the build’s context window, output limit and two prices. Enter those once and an uncensored build reads your repository, edits files and runs commands inside VS Code, on one model for Plan and another for Act if you choose to split them. The same ids also work in Cursor, in OpenCode and, with a router in front, in Claude Code.
What you need
- A key from the console. It begins
sk-unr-and one key covers every model on the page. - Cline in VS Code or the
clineCLI. Both read~/.cline/data/settings/providers.json, so a provider set up in one is set up in the other. - One id from the models page, typed as the slug.
- The build’s context window, output limit and two prices, from the tables lower down.
Not sure the model suits you yet? The same builds answer in the browser at /chat, three free turns a day with no account, and on Telegram.
The shortest path to a first reply
- Click the gear (⚙️) in the Cline panel’s top bar. The provider settings open: an API Provider dropdown with the fields for the chosen provider beneath it.
- Set API Provider to OpenAI Compatible. That is Cline’s entry for any server that speaks the OpenAI chat format from its own base URL. Skip the plain OpenAI entry; it is a different provider with a different settings page.
- Base URL:
https://api.unrestrictable.ai/v1. Keep the/v1and put nothing after it. Cline appends/chat/completionson its own. - API Key: paste the full key,
sk-unr-included. - Model ID: a single id, spelled the way the models page spells it. Cline users on this API reach for these three most; the top one is what the rest of this guide assumes:
glm-5.3-uncensoredglm-5.3-flash-uncensoreddeepseek-v4.1-flash-abliterated- Expand Model Configuration beneath the model field and enter the build’s numbers from the table in the next section. Context Window and Max Output Tokens go in as the catalog publishes them. Image Support stays off until the model page says the build takes image input. Computer Use goes on only for a build whose Tool calls column reads yes. Input Price and Output Price are the price table’s per-million rates; with those in place the cost Cline prints under a task agrees with the console.
- Press Verify. Cline fires one request at the base URL with your key and id and reports whether it came back.
- Optional: in Cline Settings, enable Use different models for Plan and Act, then pick a model per mode. The common pairing is a fast build like
glm-5.3-flash-uncensoredon Plan and a strong one likeglm-5.3-uncensoredon Act, and Cline changes model when you change mode.
Working from a terminal instead? One CLI command writes provider, key, model and base URL together into the shared file:
cline auth -p openai-compatible -k "sk-unr-live-…" -m glm-5.3-uncensored -b https://api.unrestrictable.ai/v1From then on cline runs on that model. Pass -m (--model) to use a different id for one run only; the VS Code extension sees whatever the CLI wrote, since both read the same settings file.
Plan, Act and why the split exists
Plan mode reads your code and talks the approach over with you; nothing is written. Act mode writes the edits, runs commands and reads their output. A Plan turn is almost all input plus a short reply, work a fast build does well. An Act turn is decided by tool calls and edit quality, where a strong build pays for its higher rate. Leave the switch off and a single model handles both.
Each setting, explained
- API Provider
- OpenAI Compatiblethe entry that accepts a base URL of its own; in the CLI and SDK its id is openai-compatible
- Base URL
- https://api.unrestrictable.ai/v1the destination of every request; it must end at /v1 with no path after it
- API Key
- the whole sk-unr- keygoes out as the bearer token on each call; lives in providers.json, never in the repository
- Model ID
- glm-5.3-uncensoredthe slug from the models page, not the display name; a provider entry holds one id at a time
- Context Window
- from the table belowthe ceiling Cline keeps the conversation under; it drops older messages before reaching it, so a wrong number drops them too soon or too late
- Max Output Tokens
- from the table belowthe most Cline asks for in one reply; an edit longer than this gets truncated
- Image Support
- off unless the model page lists image inputallows screenshots and pasted images in a task
- Computer Use
- only when Tool calls is yesCline’s docs keep this for models that do tool or function calling
- Input Price / Output Price
- per million tokens, from the price tablethe basis of the cost Cline displays per task; billing happens at the catalog rate whatever you type here
- Use different models for Plan and Act
- optionala model per mode; Cline swaps as the mode changes and remembers both choices
Which build to put in the field
Cline speaks the OpenAI format and each build on the models page replies in it. Most of what an agent turn costs is input: the whole conversation, every file Cline has opened and every tool result so far travel with each request. A low input price is therefore what makes a turn cheap. Act mode needs tool calls to edit and to run commands, so the Tool calls column tells you which builds can take Act; the last two columns are the two numbers Model Configuration wants.
| Build | Cost of one turn | Tool calls | Context Window | Max Output Tokens |
|---|---|---|---|---|
| GLM-5.3 Uncensored | $0.01 | model page | model page | model page |
| GLM-5.3 Flash Uncensored | $0.0047 | model page | model page | model page |
| DeepSeek V4.1 Flash Abliterated | $0.04 | model page | model page | model page |
| DeepSeek V4 Pro Uncensored | $0.04 | model page | model page | model page |
| GLM-5.3 Abliterated | $0.09 | model page | model page | model page |
| GLM-5.3 Flash Code Uncensored | $0.25 | model page | model page | model page |
| GLM-5.3 Uncensored | Uncensored | $0.282 | $0.846 |
| GLM-5.3 Flash Uncensored | Uncensored | $0.1053 | $0.3685 |
| DeepSeek V4.1 Flash Abliterated | Abliterated | $0.8488 | $1.6975 |
| DeepSeek V4 Pro Uncensored | Uncensored | $0.97 | $1.94 |
| GLM-5.3 Abliterated | Abliterated | $2.1825 | $4.365 |
| GLM-5.3 Flash Code Uncensored | Uncensored | $6.00 | $6.00 |
A real session costs less than one turn times the turn count, for two reasons. The first is caching. You pay per token at the model’s rate from a prepaid balance, and any input tokens a response reports as cached are billed at the cached rate, not the input rate. Every Cline turn resends the same system prompt, the same rules and the same opening exchange, so on a long task most of the input is that repeated prefix. The second is the Plan and Act split, which hands the read-only turns to the fast build. The two figures Model Configuration wants, input and output per million tokens, are in the price table; the cached rate sits beside them on each model page.
If the work is code, the uncensored coding AI guide lines up the code builds and prices a whole day of agent work on each.
Prove the connection
Ask the API which models your key may call. The id you put in Model ID should be in the list:
curl https://api.unrestrictable.ai/v1/models -H "Authorization: Bearer sk-unr-live-…"Next, send the kind of request Cline sends. The -i flag shows the response headers: x-uplink-request-id identifies the call, and when the answer is not streamed x-uplink-charge states its cost in USD.
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."}]}'Then run a task in Cline with one of your ids selected. It appears on the console’s usage page with its token counts and its charge, and the figure Cline shows under the task equals it once Model Configuration carries the catalog’s two prices.
When something fails
401 invalid_api_key, shown by Cline as Invalid API Key: the key is mistyped, revoked, or carries a stray space. Paste a fresh copy from the console into API Key.404 model_not_found, Cline’s Model Not Found: no such id on the models page. Compare the spelling with the page; the id is the slug, never the display name.404on every call even with a correct id: the Base URL carries/chat/completionsor some other path after/v1. Cline adds the route itself, so the field must stop at/v1.402 insufficient_balance: the request costs more than the balance holds. Top up in the console.402 monthly_limit_reachedmeans the key reached the monthly spend cap you gave it.403 scope_denied: the key is restricted to particular models and this id is outside that set.429 rate_limited: the key’s rate limit is hit. Readretry-afteron the response.- Connection Errors with no status code at all: a typo in the Base URL, or a proxy or firewall between VS Code and the API. Cline’s networking page explains its proxy settings, and the first curl above tells you whether the machine can reach the API in the first place.
- Act mode keeps saying You did not use a tool in your previous response: the build replied in prose where Cline wanted a tool call. Move Act to a build whose Tool calls column reads yes and leave the prose-only builds on Plan.
- The start of a long task vanishes: as the window fills, Cline trims the oldest messages, and a Context Window value below the build’s real one makes that trimming start early. Enter the number from the table. When the API reports the window exceeded, Cline also retries with a shorter conversation.
Questions
https://api.unrestrictable.ai/v1, ending at /v1. Cline appends /chat/completions to it on every request, so a Base URL that already includes the route gets a 404 back.
The slug as the models page prints it, such as the first id listed under the steps above; never the display name. Cline forwards the string as typed, so a misspelling returns Model Not Found. The first curl in the connection check prints every id your key can use. See the models page.
One from this console, beginning sk-unr-, pasted whole into the API Key field or handed to cline auth with -k. A single key reaches every model on the page; Cline keeps it in providers.json, which the extension, the CLI and the SDK all read.
You pay per token at the model’s rate out of a prepaid balance; there is no subscription. The table above prices one agent-sized turn on each build at the live rate, and once the catalog prices are in Model Configuration, the number Cline prints under a task is the number the console bills.
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.