Docs

Msty Go

Msty Go's provider form is five fields, and its own documentation names all five. It does not name what to put in two of them — so this page says which two, and how to settle each one in under a minute.

Settings > Model Providers takes a Base URL and an API key — five documented fields, two of which Msty Go's own docs leave you to settle by test.

Settings > Model Providers

Settings > Model Providers  —  add a provider

Type                      an OpenAI-compatible provider type
                          (Go's docs name the field, not its options)
Display name              Kunavo
Base URL                  https://api.kunavo.com/v1     <- read the note
API key (when required)   sk-kn-...
Model list behavior       if Go does not fetch the list, type ids by hand

Custom model entry (non-managed providers)
  Context window tokens   optional
  Reserved output tokens  optional

The base URL form is not settled by Msty Go's documentation. No page under docs.msty.ai/go gives an example base URL or states whether Go appends /v1 itself, so this page will not pretend to know. Start with https://api.kunavo.com/v1, which is Kunavo's OpenAI-compatible root; if the model list comes back empty or the first message 404s while the curl below returns JSON, the endpoint and key are fine and the remaining suspect is the suffix — retry with the bare origin https://api.kunavo.com. One of the two works and the test costs a request.

Type is the same shape of gap. Msty Go's Model Providers page names the field and never enumerates its options. The first-party evidence that an OpenAI-compatible provider type exists is Msty's own changelog, which records fixes to “OpenAI-compatible providers” in 0.15.0 (August 6, 2026) and to OpenAI-compatible models in 0.3.0 (April 22, 2026). Pick that type; the exact label in your build is the authority, not this page.

Kunavo has not run Msty Go against this endpoint. Go is a closed-source beta desktop app that nobody here has installed, and a published setup page is not a test. Everything above was read from Msty's own documentation and changelog on the date below. Streaming, tool calling and vision behaviour through a custom provider are not enumerated in Msty Go's documentation either — that is a gap in the docs, not a claim that they are missing.

Step by step

Create a key at /app/keys and copy it — it is shown once.
Open Settings > Model Providers and add a provider. The documented fields are Type, Display name, Base URL, API key (when required) and Model list behavior, in that order.
If no model list appears, add the ids by hand — the docs say “For non-managed providers, you can add custom model entries manually”, with Context window tokens and Reserved output tokens as optional metadata. Kunavo also answers GET /v1/models, so an empty list usually points at the base URL rather than at the key.
Do what the docs tell you to do next: “After setup, run at least one realistic prompt through the provider before assigning it to agents or scheduled automation.” Go's own minimum checks are that you can select at least one model and that a test chat response succeeds.
Open Settings > Model Assignments and point the background targets — Memory, Context Scout, Context Summary, Conversation Titles, Side Replies (/btw) and Pulse — at a cheap model, keeping the expensive one for chat. Pulse will not run until it has an explicit Pulse model.

Checked against Msty Go's Model Providers page on September 21, 2026. Third-party settings move; if a field name here no longer matches what you see, that page is the authority, not this one.

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 Go.

# 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 Go
`claude-sonnet-5`	$2.00 / $10.00	The agent's working model — the one chat and playbook runs should inherit
`claude-opus-5`	$2.00 / $10.00	Mission Control planning and handoffs, where a wrong plan costs a whole run
`claude-haiku-4-5`	$0.40 / $2.00	Model Assignments: Memory, Conversation Titles, Side Replies — volume, not depth
`gemini-3-7-flash`	$0.525 / $2.625	Pulse and Context Scout, which run unattended and on a schedule

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.

Five things answer to “Msty”. This page is one of them

Getting this wrong is the most common way to end up configuring the wrong app, or quoting the wrong price.

Msty Go — the desktop agent app this page is about. It was renamed from Msty Claw in release 0.14.0 on July 26, 2026, and its documentation still serves from a Claw-era host, so old links and old names are still in circulation. Follow them; they land here.
Msty Studio — a different product with a different provider screen (Model Hub > Model Providers) and its own prices. Its setup is not interchangeable with Go's.
Msty (v1.x), retired, whose frozen docs still rank; Msty Nexus and Msty Stack, siblings with their own pricing.
MSTY on NYSE Arca — an exchange-traded fund. Its share price is not a software price and has nothing to do with any of the above.

What Go costs today, and what that does not cover

Msty's pricing page, read September 21, 2026, puts Go in a row marked Free / Beta access and says: “Go is the new name for Claw. Free beta access is available now; team plans are coming soon.” That is the application only. A free client does not make the model calls free — those are billed by whoever issued the key in the Base URL field, which on this page is Kunavo, per token from a prepaid balance. Read Msty's current terms yourself before putting a beta app into commercial work; the beta's permitted-use wording is theirs to change and this page does not restate it.

One Go feature this endpoint is not involved in

Release 0.15.0 added local speech-to-text with downloadable on-device models. That runs on the machine and reaches no provider at all, which is just as well: Kunavo serves no speech-to-text model and no embedding model, so nothing in that feature is billable here and nothing in it is a reason to pick a provider.

For the rest, see Msty Go's changelog, which is the only first-party place the OpenAI-compatible provider type is named at all.

FAQ

How do I add a custom API provider to Msty Go?

Open Settings > Model Providers and add a provider. Msty Go's documentation lists the fields as Type, Display name, Base URL, API key (when required) and Model list behavior, in that order, and adds that for non-managed providers you can enter custom model entries manually with optional context-window and reserved-output metadata. The documentation names the Type field but does not publish its options, so the exact wording of the OpenAI-compatible choice is whatever your build shows. After saving, Msty's own instruction is to run at least one realistic prompt through the provider before assigning it to agents or scheduled automation.

What base URL does Msty Go need — with or without /v1?

Msty Go's documentation does not answer this. No page under docs.msty.ai/go gives an example base URL or states whether Go appends /v1 itself, so anyone telling you definitively is guessing. Settle it by test rather than by trust: enter https://api.kunavo.com/v1 first, because that is the OpenAI-compatible root the endpoint publishes, and if the model list stays empty or requests 404 while a plain curl to https://api.kunavo.com/v1/models returns JSON, change the field to the bare origin https://api.kunavo.com and try again. The curl separates an endpoint or key problem from a suffix problem in one request.

Where does Msty Go's API key go, and what is it billing?

Into the provider's "API key (when required)" field under Settings > Model Providers, next to the Base URL it authenticates. That pairing is the whole billing relationship: the key belongs to whoever runs the endpoint you typed, and every chat, agent run, playbook and scheduled task that resolves through that provider is charged there. Msty Go itself is free during its beta according to Msty's pricing page, which means the application costs nothing and the model calls are a separate, metered bill from the provider.

Can Msty Go use a cheap model for background work and an expensive one for chat?

Yes, and it is the single largest cost lever in the app. Settings > Model Assignments exposes per-task targets — Memory, Context Scout, Context Summary, Conversation Titles, Side Replies (/btw) and Pulse — and Msty's own documentation recommends low-cost or faster models there while keeping stronger models for core chat execution. Pulse additionally requires an explicit Pulse model plus a visible provider that can run it before it will start. These background jobs run unattended and on a schedule, so a flagship model assigned to Conversation Titles bills continuously for work nobody reads closely.

Has Kunavo tested Msty Go?

No. Msty Go is a closed-source beta desktop application and nobody at Kunavo has installed it, so nothing on this page is a runtime result. The field names, their order, the manual-model path and the Model Assignments targets were read from Msty's own documentation on September 21, 2026, and the existence of an OpenAI-compatible provider type comes from Msty's own changelog. Treat a published setup page as a reading of someone else's documentation, which is what it is, and treat your own first successful request as the actual confirmation.