tokmine: command not found after installing.

Open a new terminal so the updated PATH is loaded. If it persists, add the directory the installer printed to your PATH, or reinstall.

The installer fails behind a corporate proxy.

Set HTTPS_PROXY (and HTTP_PROXY) in the shell running the installer. The CLI honours the same variables at runtime.

The download is corrupt or was altered in transit, so nothing was installed. Retry; if it repeats, contact us.

The CLI finds no models.

Make sure the runner is running and listening on localhost. Run tokmine models to see what was discovered, or pass --provider http://host:port for a non-default address. See Supported model runners.

"No local model server found" when I start tokmine.

No runner answered on the default ports. Run tokmine install ollama to install one and pull a model that fits your machine, or start your own runner. In an interactive terminal the CLI offers to do it for you. See Install a local model.

tokmine install refuses to run.

Without a terminal it needs --yes to confirm the plan, or use --dry-run to only print it. With no runner argument in a non-interactive shell it prints the options and exits with code 2.

"Only one anonymous session per IP" (HTTP 429).

Another tokmine process, or a recent session from the same network, is still active. Stop it, wait for it to expire, or run tokmine login for a permanent tunnel.

"The agent token was rejected".

The stored token is wrong or revoked. Run tokmine login with a fresh token from the portal, or tokmine logout to run anonymously.

"Plan limit reached".

You have more connected machines than your plan allows. Disconnect one or change plan in the portal.

"Endpoint not found" or "endpoint disabled".

The slug does not exist in your workspace, or the endpoint was disabled or deleted, so the CLI exits instead of reconnecting. Run tokmine endpoints to see the valid slugs and restart with --endpoint . Slugs cannot be changed after creation.

"Plan required" when creating or joining an endpoint.

Static endpoints need a paid plan, or you reached your plan's endpoint limit. Free accounts keep the default endpoint. Upgrade in the portal, or remove an endpoint you no longer use.

The CLI says I have several endpoints and exits.

In a script or service there is no terminal to ask you. Pass --endpoint (or set TOKMINE_ENDPOINT); tokmine endpoints lists them. In a terminal it shows a picker and remembers the choice.

The error says I passed a URL to --endpoint.

--endpoint now takes an endpoint slug. The server URL moved to --server (TOKMINE_SERVER). A URL in --endpoint still works for now, with a deprecation notice.

"Connected from another tokmine process".

Two processes share the same machine id. Only one may run per machine; the CLI exits to avoid a reconnect fight.

"This CLI is too old".

Re-run the install command to upgrade.

What happens if my network drops or the laptop sleeps?

The CLI reconnects automatically with exponential backoff (1 to 30 seconds). Anonymous sessions resume with the same URL and token. Ctrl-C drains in-flight requests before exiting.

I get 401 from the public URL.

Send the consumer token printed by the CLI (anonymous) or your own account token (authenticated). Any of Authorization: Bearer, x-api-key, x-goog-api-key or x-token works.

I get 404 or a model-not-found error.

The model name must match one reported by tokmine models exactly, for example llama3.2 or llama3.2:latest.

The tunnel is up but the local runner refused or dropped the request, or, for an endpoint, no healthy machine is online in its pool. Check that the runner and the CLI are still running.

I get 404 from a static endpoint URL.

The endpoint may be disabled or deleted, or no machine of its pool serves that model (the error names which). Check the Endpoints page and tokmine models.

Anonymous sessions are rate limited. Log in for higher limits.

The chat says my machine is offline.

The tunnel is not connected. Check that tokmine is still running (anonymous sessions end after 30 minutes) and that your runner is up, then reload the chat.

There is a token in the chat link. Is that safe?

It sits in the URL fragment, which is not sent to servers. The page verifies it, moves it into an httpOnly, Secure, SameSite=Strict cookie and clears the address bar. The cookie lasts 30 minutes by default (12 hours at most). Treat the link like a password and do not share it.

The chat stopped working after 30 minutes.

The anonymous session expired. Restart tokmine to get a new link, or log in for a persistent machine.

Do I need to open ports or configure my router?

No. The CLI opens an outbound TLS connection to Tokmine. Nothing listens on the public internet from your machine.

What happens when the free session ends?

After 30 minutes the tunnel closes and the URL is freed. Log in with an agent token to keep a persistent machine.

Can I use it from Docker or a server?

Yes. Set TOKMINE_TOKEN in the environment and run tokmine. Point --provider at the model container.

Docs

Troubleshooting and FAQ

Installing

tokmine: command not found after installing.
Open a new terminal so the updated PATH is loaded. If it persists, add the directory the installer printed to your PATH, or reinstall.
The installer fails behind a corporate proxy.
Set HTTPS_PROXY (and HTTP_PROXY) in the shell running the installer. The CLI honours the same variables at runtime.
Checksum mismatch.
The download is corrupt or was altered in transit, so nothing was installed. Retry; if it repeats, contact us.

Running the CLI

The CLI finds no models.
Make sure the runner is running and listening on localhost. Run tokmine models to see what was discovered, or pass --provider http://host:port for a non-default address. See Supported model runners.
"No local model server found" when I start tokmine.
No runner answered on the default ports. Run tokmine install ollama to install one and pull a model that fits your machine, or start your own runner. In an interactive terminal the CLI offers to do it for you. See Install a local model.
tokmine install refuses to run.
Without a terminal it needs --yes to confirm the plan, or use --dry-run to only print it. With no runner argument in a non-interactive shell it prints the options and exits with code 2.
"Only one anonymous session per IP" (HTTP 429).
Another tokmine process, or a recent session from the same network, is still active. Stop it, wait for it to expire, or run tokmine login for a permanent tunnel.
"The agent token was rejected".
The stored token is wrong or revoked. Run tokmine login with a fresh token from the portal, or tokmine logout to run anonymously.
"Plan limit reached".
You have more connected machines than your plan allows. Disconnect one or change plan in the portal.
"Endpoint not found" or "endpoint disabled".
The slug does not exist in your workspace, or the endpoint was disabled or deleted, so the CLI exits instead of reconnecting. Run tokmine endpoints to see the valid slugs and restart with --endpoint <slug>. Slugs cannot be changed after creation.
"Plan required" when creating or joining an endpoint.
Static endpoints need a paid plan, or you reached your plan's endpoint limit. Free accounts keep the default endpoint. Upgrade in the portal, or remove an endpoint you no longer use.
The CLI says I have several endpoints and exits.
In a script or service there is no terminal to ask you. Pass --endpoint <slug> (or set TOKMINE_ENDPOINT); tokmine endpoints lists them. In a terminal it shows a picker and remembers the choice.
The error says I passed a URL to --endpoint.
--endpoint now takes an endpoint slug. The server URL moved to --server (TOKMINE_SERVER). A URL in --endpoint still works for now, with a deprecation notice.
"Connected from another tokmine process".
Two processes share the same machine id. Only one may run per machine; the CLI exits to avoid a reconnect fight.
"This CLI is too old".
Re-run the install command to upgrade.
What happens if my network drops or the laptop sleeps?
The CLI reconnects automatically with exponential backoff (1 to 30 seconds). Anonymous sessions resume with the same URL and token. Ctrl-C drains in-flight requests before exiting.

Calling the URL

I get 401 from the public URL.
Send the consumer token printed by the CLI (anonymous) or your own account token (authenticated). Any of Authorization: Bearer, x-api-key, x-goog-api-key or x-token works.
I get 404 or a model-not-found error.
The model name must match one reported by tokmine models exactly, for example llama3.2 or llama3.2:latest.
I get 502.
The tunnel is up but the local runner refused or dropped the request, or, for an endpoint, no healthy machine is online in its pool. Check that the runner and the CLI are still running.
I get 404 from a static endpoint URL.
The endpoint may be disabled or deleted, or no machine of its pool serves that model (the error names which). Check the Endpoints page and tokmine models.
I get 429.
Anonymous sessions are rate limited. Log in for higher limits.

Browser chat

The chat says my machine is offline.
The tunnel is not connected. Check that tokmine is still running (anonymous sessions end after 30 minutes) and that your runner is up, then reload the chat.
There is a token in the chat link. Is that safe?
It sits in the URL fragment, which is not sent to servers. The page verifies it, moves it into an httpOnly, Secure, SameSite=Strict cookie and clears the address bar. The cookie lasts 30 minutes by default (12 hours at most). Treat the link like a password and do not share it.
The chat stopped working after 30 minutes.
The anonymous session expired. Restart tokmine to get a new link, or log in for a persistent machine.

See Test your endpoint in the browser.

General

Do I need to open ports or configure my router?
No. The CLI opens an outbound TLS connection to Tokmine. Nothing listens on the public internet from your machine.
What happens when the free session ends?
After 30 minutes the tunnel closes and the URL is freed. Log in with an agent token to keep a persistent machine.
Can I use it from Docker or a server?
Yes. Set TOKMINE_TOKEN in the environment and run tokmine. Point --provider at the model container.

Still stuck? Run tokmine status --verbose and contact us. Tokens are never printed in logs.