Docs

Msty Studio

Msty Studio's documentation says an OpenAI-compatible endpoint can be added as a provider and then stops. The app itself is more specific — it names the provider type, the two fields and the /v1 rule. Here is that form, filled in for Kunavo.

Model Hub > Model Providers > Add Provider > Open AI Compatible: an API Key and an Inference Endpoint of https://api.kunavo.com/v1 — the app's own note says most OpenAI-compatible endpoints need the /v1.

Msty Studio 2.9.11 — Add Provider

Model Hub  >  Model Providers  >  Add Provider

Provider             Open AI Compatible     (under "OpenAI Compatible")
Name                 Kunavo                 "Give it a name (optional)"
API Key              sk-kn-...
Inference Endpoint   https://api.kunavo.com/v1      <- keep the /v1

Models               Fetch Models, then tick the ids you want
                     (or Add Custom Model: Model ID + Model Name)
                                                     → Save Provider

Keep the /v1 — Msty says so itself. Under the Inference Endpoint field, the shipped app prints: “Most OpenAI-compatible provider endpoints require the /v1 path, e.g. https://example.com/v1.” Msty Studio adds only /models and /chat/completions to what you type. Without /v1, Fetch Models asks https://api.kunavo.com/models — a web page, not a JSON list — and chat requests land on a 404.

Where these labels come from. Msty's online documentation does not name the fields. The labels above are the strings shipped in the Msty Studio 2.9.11 desktop build (downloaded from Msty's own servers on October 1, 2026 and read without launching it), so they match what the app displays in English. A later release may reword them; the app on your screen is the authority.

Not run end to end. Nobody here has sent a message from Msty Studio to Kunavo. The provider form, the /v1 note and the request paths were read from the released build, and Kunavo's side was checked with plain HTTP requests. Streaming, tool calls, image input and thinking controls through this provider type are untested — your first chat is the real check.

Step by step

Create a key at /app/keys and copy it — it is shown once.
In Msty Studio open Model Hub > Model Providers and choose Add Provider. In the provider list, under the OpenAI Compatible heading, pick Open AI Compatible — not OpenAI, which is the fixed OpenAI provider.
Give it a name, paste the key into API Key, and enter https://api.kunavo.com/v1 as the Inference Endpoint. The app requires it to start with https://.
Click Fetch Models and tick the models you want. Kunavo answers GET /v1/models with the full catalog, so select a few rather than all. If you prefer, open Add Custom Model and type a Model ID exactly as the catalog spells it, for example claude-sonnet-5.
Save Provider, then open a conversation and pick one of its models — Msty's own docs give that as the test step. If the reply arrives, the endpoint, key and model id are all right.

Checked against Msty Studio's Online Providers page on October 1, 2026. Third-party settings move; if a field name here no longer matches what you see, that page is the authority, not this one.

This is the short version. The full walkthrough — model choice, what a real session costs, and the failure modes — is in Msty Studio pricing and API cost.

Verify before you debug the client

One request settles whether a failure is the endpoint, the key, or the configuration file. If this returns JSON, the same base URL and key work in Msty Studio.

# Settles whether a failure is the endpoint, the key, or the client.
curl -sS https://api.kunavo.com/v1/models \
  -H "Authorization: Bearer sk-kn-..."

Which model id to put in the field

Every text model is reachable as a model id — the live list is GET /v1/models, and the catalog with prices is on the models page. Rates are USD per 1M tokens, input / output.

Model id	Kunavo in / out	Where it fits in Msty Studio
`claude-sonnet-5`	$1.40 / $7.00	Long documents and Knowledge Stack chats, the model most conversations should default to
`gpt-5-6-terra`	$0.70 / $4.20	A second family in a split-chat comparison pane
`claude-haiku-4-5`	$0.70 / $3.50	Quick questions, titles and anything you run many times

Billing is per token from a prepaid balance with no monthly fee — see billing. On repeated context — which is most of what an editor or a chat client sends — prompt caching moves the bill more than the model choice does.

Which Msty this is

Msty Studio — the desktop and web workspace this page is about, with providers under Model Hub > Model Providers.
Msty Go — a separate agent app with its own provider screen under Settings; its setup is on its own page.
Msty 1.x — the retired app, whose frozen docs at docs.msty.app still rank. Its screens are not Studio's.

What the app tells you when it fails

These are the messages shipped in 2.9.11, and what each one means for this endpoint:

“The provider endpoint was not found. Verify the endpoint URL and make sure it includes the correct API path, such as /v1.” — the Inference Endpoint is missing /v1.
“Unable to connect to the provider endpoint…” — a typo in the host, or http:// where it should be https://.
Fetch Models fails but the URL is right — check the key. Kunavo's model list needs a valid key and answers 401 without one.
A model is listed but a chat errors — Msty's own form warns that “Not all models have been tested with Msty”. Kunavo's list includes image, video and audio models that a chat window cannot use; pick a chat model.

What leaves your machine

Msty describes online providers as a way to “use hosted models with your API keys, without running model inference on your device”. So a conversation on a Kunavo model sends that conversation's prompt and attached context to Kunavo, which forwards it to the model vendor; a conversation on a local model does not. Keep sensitive material on local models if that boundary matters to you.

Thinking effort on this provider type

Msty's documentation lists thinking-effort support for “OpenAI Compatible Models” as None and Default only, while its native Anthropic provider gets Think Lightly, Moderately and Deeply. Through the Open AI Compatible type, expect a Claude model to run at its default reasoning rather than the graded levels.

FAQ

How do I add a custom OpenAI-compatible provider in Msty Studio?

Open Model Hub > Model Providers and choose Add Provider. Under the OpenAI Compatible heading pick Open AI Compatible, give it a name, paste your key into API Key, and enter the endpoint in Inference Endpoint including the version path — for Kunavo, https://api.kunavo.com/v1. Click Fetch Models to load the list, or use Add Custom Model to type a Model ID, then Save Provider. Those labels are the strings in the Msty Studio 2.9.11 build; Msty's online documentation describes the same flow without naming the fields.

Does Msty Studio need /v1 in the endpoint?

For most OpenAI-compatible providers, yes, and the app says so itself: under the Inference Endpoint field it shows "Most OpenAI-compatible provider endpoints require the /v1 path, e.g. https://example.com/v1. If you are running into issues, try appending /v1 to the endpoint." Msty Studio appends only /models and /chat/completions, so Kunavo's endpoint is https://api.kunavo.com/v1. Without /v1 the model fetch reads a web page instead of a model list and chats return a 404.

Is Msty Studio's custom provider the same as Msty Go's?

No. They are separate apps with separate provider screens. Msty Studio adds providers under Model Hub > Model Providers with the Open AI Compatible type described here. Msty Go, the agent app formerly called Msty Claw, configures providers under Settings > Model Providers with its own set of fields. Setup instructions for one do not transfer to the other, and neither matches the retired Msty 1.x app whose old documentation still appears in search results.

Who bills me when Msty Studio uses a custom provider?

Two parties, separately. Msty bills for the app tier you use, if any. The model calls are billed by whoever issued the key in the provider's API Key field — with a Kunavo key, per token from a prepaid Kunavo balance. Local models you run inside Msty cost nothing per request. The tiers and what a typical long-document conversation costs at catalog rates are worked through on the Msty Studio pricing page linked above.

Has Kunavo tested Msty Studio?

Not end to end. On October 1, 2026 the official Msty Studio 2.9.11 desktop build was downloaded from Msty's servers and read without being launched: the provider type, field labels, the /v1 note, the error messages and the request paths on this page come from it. Kunavo's side was checked with direct HTTP requests. No chat from Msty Studio to Kunavo has been run, so streaming, tool use and image input through this provider are unverified.