Continue

Continue is a coding assistant for VS Code and JetBrains. Its configuration file supports provider: openai, apiBase, apiKey, and model, so it can connect to an OpenAI-compatible endpoint.

Preparation

Prepare three pieces of information first:

Field	Value
Base URL	https://n.tokeness.dev/v1
API Key	A key created in the Tokeness console
Model	Full model name copied from the Tokeness model marketplace

If you want Continue to read and write files in Agent mode, the model needs to support tool calling. On first configuration, start with a regular chat to confirm the key, Base URL, and model name are correct, then test Agent mode.

Open the configuration file

Open the configuration file from the Continue panel. The entry point name varies by version, but common locations are:

1. Open VS Code or JetBrains.
2. Open the Continue extension panel.
3. Go to settings or the configuration entry.
4. Open the local config.yaml.

You can also look for the Continue configuration directory under your user directory, for example ~/.continue/config.yaml.

Basic configuration

Add the following to models. Replace YOUR_TOKENESS_API_KEY and YOUR_MODEL_NAME with your own values.

name: Tokeness Local
version: 1.0.0
schema: v1

models:
  - name: Tokeness Chat
    provider: openai
    model: YOUR_MODEL_NAME
    apiBase: https://n.tokeness.dev/v1
    apiKey: YOUR_TOKENESS_API_KEY
    roles:
      - chat
      - edit
      - apply

Field meanings:

Field	Description
name	The name displayed in the Continue interface, e.g. Tokeness Chat
provider	Set to openai so Continue uses the OpenAI-compatible calling method
model	Full model name from the Tokeness model marketplace
apiBase	Set to https://n.tokeness.dev/v1
apiKey	Tokeness API key
roles	The tasks this model can handle

After saving, restart the IDE or refresh the configuration in the Continue panel.

Store the key in an environment variable

If you do not want to write the key into the configuration file, put it in an environment variable first.

macOS / Linux:

export TOKENESS_API_KEY="your Tokeness API key"

Windows PowerShell:

$env:TOKENESS_API_KEY = "your Tokeness API key"

Then reference it in the configuration:

models:
  - name: Tokeness Chat
    provider: openai
    model: YOUR_MODEL_NAME
    apiBase: https://n.tokeness.dev/v1
    apiKey: ${{ env.TOKENESS_API_KEY }}
    roles:
      - chat
      - edit
      - apply

If your Continue version does not recognize this syntax, test with the local configuration file first, and once it works, adjust according to the secret management method supported by your current Continue version.

Agent mode

Agent mode delegates file reads, file modifications, and command execution to the model for scheduling. You can explicitly declare tool-calling capability in the configuration:

models:
  - name: Tokeness Agent
    provider: openai
    model: YOUR_TOOL_CALLING_MODEL
    apiBase: https://n.tokeness.dev/v1
    apiKey: YOUR_TOKENESS_API_KEY
    roles:
      - chat
      - edit
      - apply
    capabilities:
      - tool_use

Here YOUR_TOOL_CALLING_MODEL should be replaced with a model confirmed to support tool calling on Tokeness. Do not judge by whether the model name looks like a coding model — use whether the actual call produces tool calls as the criterion.

Verification

After saving the configuration, verify in the following order:

1. Select Tokeness Chat in Continue.
2. Send a simple question, such as "reply with ok only".
3. Confirm a reply appears in the IDE.
4. Open the Tokeness usage logs and confirm a corresponding request appears.
5. Then test code explanation, code rewriting, or Agent mode.

If chat works but rewriting or Agent mode is unstable, the issue is usually not the key — it is model capability, tool-calling format, or Continue version differences.

Common issues

Symptom	Fix
Model does not appear in the dropdown	Check config.yaml indentation, schema, and models structure; restart the IDE after saving
401	Re-copy the Tokeness API key and confirm there are no extra spaces
404	Check that apiBase is https://n.tokeness.dev/v1
model not found	Re-copy the model name from the Tokeness model marketplace
Agent tool calls fail	Switch to a model that supports tool calling, or use only chat, explanation, and rewriting first
Configuration looks correct but still uses the old model	Close all IDE windows and reopen

References

• Continue OpenAI configuration
• Continue config.yaml Reference