Continue

Continue is an open-source AI code assistant for VS Code and JetBrains IDEs: chat, edit, and agent modes driven by your own model endpoints. Add Nebul as an openai provider in its config and every role runs on your project's models.

Prerequisites

• VS Code or a JetBrains IDE
• A Nebul AI Studio account with API access
• An API key for your active project
• A model ID from the Model Catalog

Installation

Install Continue from the VS Code Marketplace or the JetBrains plugin repository, then open the sidebar with cmd/ctrl+L in VS Code.

Connect it to Nebul

Continue reads models from ~/.continue/config.yaml (Windows: %USERPROFILE%\.continue\config.yaml). Open it from the sidebar's config dropdown, the cog next to Local Config, and add a Nebul model:

name: Local Config
version: 1.0.0
schema: v1
models:
  - name: GLM 5.3
    provider: openai
    model: zai-org/GLM-5.3
    apiBase: https://api.inference.nebul.io/v1
    apiKey: sk-your-api-key-here
    roles:
      - chat
      - edit
      - apply
    capabilities:
      - tool_use
    defaultCompletionOptions:
      contextLength: 1048576
      maxTokens: 16384

Save the file. Continue reloads config automatically, and the model appears in the selector above the chat input.

What the fields do:

• roles assigns the model to chat, edits, and applying edits. Narrow the list if you want different models per role.
• capabilities: [tool_use] is required for agent mode on custom endpoints. Continue can't always autodetect tool calling through a custom base URL, so declare it explicitly.
• contextLength and maxTokens shape context management. Take contextLength from /v1/model/info; a conservative output cap like the one above is fine.

Rather not put the key in the file? Omit apiKey and Continue prompts for it in the UI, storing it via the OS keychain.

A second model for autocomplete

Autocomplete works best with a small, fast model:

  - name: Qwen3 30B
    provider: openai
    model: Qwen/Qwen3-30B-A3B-Instruct-2507
    apiBase: https://api.inference.nebul.io/v1
    apiKey: sk-your-api-key-here
    roles:
      - autocomplete
    autocompleteOptions:
      debounceDelay: 250
      maxPromptTokens: 1024

Troubleshooting

• Model not in the selector: check the config dropdown at the top-right of the chat input. Continue may still be pointed at a hub config; switch to Local Config.
• Agent mode refuses to run tools: the model entry needs capabilities: [tool_use] and a model that supports tool calling. Verify on the catalog.
• 404 on requests: apiBase must be https://api.inference.nebul.io/v1, with the /v1 suffix.
• Edits misbehave on long files: the contextLength is probably wrong. Set it from /v1/model/info rather than the default.

Data and telemetry

Continue sends anonymous telemetry to PostHog unless you turn it off. Disable it in either place:

• Continue's user settings, by turning off Allow Anonymous Telemetry; or
• VS Code settings, by unchecking Continue: Telemetry Enabled. This one overrides the settings page.

See Continue's offline guide for details. Requests sent to Nebul are covered by Privacy & Security.