How do I use Cline with FreeModel? Four VS Code settings

Setup

{
  "cline.apiProvider": "openai",
  "cline.openaiBaseUrl": "https://freemodel.online/v1",
  "cline.openaiApiKey": "sk-your-key",
  "cline.model": "fm-v1-standard"
}

The same three fields are in Cline's own settings panel if you would rather click than edit JSON. Provider: OpenAI Compatible.

Why an alias works well here

Cline is an agent — it reads files, writes patches and runs commands in a loop. That means many calls per task, and the model that is best for the planning step is often not the best for the mechanical edits.

Point it at a tier alias such as fm-v1-standard and the router makes that call per request. If a provider hits a rate limit mid-task, the request moves to the next candidate rather than failing the run.

For long agent runs specifically, a tier alias beats a pinned id: it is re-resolved on every call, so a provider that starts failing loses the next request instead of the run.

If it does not work

Symptom	Cause
404	Base URL is not ending at `/v1`, or has an extra segment
401	Key wrong, or the provider is still set to something other than OpenAI Compatible
Runs stall on long tasks	Provider-side rate limits; a route with failover handles it better than a fixed model
Tool calls rejected	The id is not the one you meant — ids are exact strings; check the spelling against `/v1/models`

Set up in three steps

Step 1: Open Cline's settings and set the API provider to OpenAI Compatible. Until that changes, the other fields are ignored and Cline keeps using whichever provider was selected before.
Step 2: Fill in the base URL and the key. The base is https://freemodel.online/v1 and the key is your sk- value from the console. Cline appends /chat/completions to whatever you put here, so /v1 is the last segment.
Step 3: Set the model id and run one small task before you run a large one. curl -s https://freemodel.online/v1/models confirms the id exists without spending a request, and a single-file edit tells you whether tool calls are reaching the model at all.

The same three fields live in VS Code's settings.json as cline.apiProvider, cline.openaiBaseUrl, cline.openaiApiKey and cline.model, if you would rather keep the configuration in a file you can commit and share with a team.

Why the model has to support tool calling

Cline does not chat, it acts. Reading a file, writing a patch and running a command are all tool calls, so a model that answers text but rejects tools breaks the loop on the first step instead of degrading gracefully. The model list does not state which ids accept tools, so the reliable move is not to pin one: send an alias and let the router re-decide on every call, where a model that fails a call is skipped rather than fatal.

The failure that costs the most is not a bad answer, it is a dead provider halfway through a task. A pinned model id has nowhere to go when that happens, while an alias moves the request to its next candidate and the loop keeps running. That is the practical reason to prefer fm-v1-standard here over naming a model you happen to like.

Cline FAQ

Cline is an agent loop, so the questions below are mostly about tool calls and about what happens when a provider dies halfway through a task. Routing the request rather than pinning an id is what keeps a long run alive when that happens: 199 listed models, 77% of them chat, each call re-decided.[1]

Do I edit settings.json, or is the panel enough?

Either. The settings panel writes the same four values, and the JSON is there for people who want the configuration in version control. There is nothing in the file that the panel cannot set.

Why does Cline ask for approval on every step?

That is Cline's own approval setting, applied before any request leaves the editor. Changing the provider does not change it, and reducing it does not make the model any safer — the gateway never sees the approval decision.

Can I switch the model partway through a task?

Yes, and with a route you do not have to. The router re-decides per request, so a provider that starts rate-limiting loses the next call rather than stalling the task. Switching the id by hand works too, but the new model starts from the conversation rather than from the state you had in mind.

What does a 404 from Cline mean?

Almost always the base URL. It has to end at /v1 with nothing after it, because Cline adds /chat/completions itself. A trailing slash or an extra segment gives you a doubled path.

Can I use a concrete model id instead of a route?

Yes, and it is the better choice when you are chasing a specific behaviour. Copy the id exactly from /v1/models — they are case-sensitive, and the auto/ prefix is part of the route names rather than decoration.

Sources

Every third-party figure quoted on this site was read from the vendor documentation linked below, on the date stated. None of these links are affiliate links and none of these vendors paid to be listed. If a number here disagrees with the page it came from, the page is right — tell us and it gets corrected.

Where the percentages come from

Every share quoted on this site is counted from the live model list, not estimated. Divide the two counts in the last column, round to the nearest whole percent, and you get the number in the first.

Share	Of what	Count
77%	are chat models	153 of 199 listed entries
9%	generate images	18 of 199 listed entries
23%	are not chat models	46 of 199 listed entries
3%	are modality routes	6 of 199 listed entries
17%	carry a context length	34 of 199 listed entries

[1] Counted on 29 September 2026 from GET https://freemodel.online/v1/models, which needs no key and no account. Every row uses all 199 listed entries as the denominator, because the only thing the endpoint states for every entry is its modality. It no longer publishes per-model capability flags such as tool calling or reasoning, so this site quotes no shares for them. Fetch the list and recount whenever you want to check the arithmetic.