Skip to main content

​
Continue.dev

Continue is open source and lets you configure custom provider endpoints. Point it at your Takumo gateway and every chat, autocomplete, and edit request is protected.

​
Setup

1

Get your API key

Log into cloud.takumo.io and copy your API key from the dashboard.
2

Open your Continue config

Continue stores its configuration at ~/.continue/config.json (or config.ts if you prefer TypeScript).Open it in your editor:
code ~/.continue/config.json
3

Set the provider endpoint

Set apiBase to your Takumo gateway URL and apiKey to your Takumo API key:
{
  "models": [
    {
      "title": "Claude via Takumo",
      "provider": "anthropic",
      "model": "claude-sonnet-4-5-20250929",
      "apiBase": "https://gateway.takumo.io/v1",
      "apiKey": "your-takumo-api-key"
    }
  ]
}
For OpenAI models:
{
  "models": [
    {
      "title": "GPT-4o via Takumo",
      "provider": "openai",
      "model": "gpt-4o",
      "apiBase": "https://gateway.takumo.io/v1",
      "apiKey": "your-takumo-api-key"
    }
  ]
}
Self-hosted gateway users: replace the URL with your own.
4

Configure autocomplete (optional)

If you use tab autocomplete, route that through Takumo too:
{
  "tabAutocompleteModel": {
    "title": "Codestral via Takumo",
    "provider": "mistral",
    "model": "codestral-latest",
    "apiBase": "https://gateway.takumo.io/v1",
    "apiKey": "your-takumo-api-key"
  }
}
5

Test it

Open a file with secrets. Ask Continue a question about the code. Check the dashboard for detections.

​
Full config example

Here’s a complete config.json with multiple models routed through Takumo: 
{
  "models": [
    {
      "title": "Claude Sonnet (Takumo)",
      "provider": "anthropic",
      "model": "claude-sonnet-4-5-20250929",
      "apiBase": "https://gateway.takumo.io/v1",
      "apiKey": "your-takumo-api-key"
    },
    {
      "title": "GPT-4o (Takumo)",
      "provider": "openai",
      "model": "gpt-4o",
      "apiBase": "https://gateway.takumo.io/v1",
      "apiKey": "your-takumo-api-key"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Codestral (Takumo)",
    "provider": "mistral",
    "model": "codestral-latest",
    "apiBase": "https://gateway.takumo.io/v1",
    "apiKey": "your-takumo-api-key"
  }
}

​
Usage

Once configured, Continue works exactly the same as before. Chat, autocomplete, and slash commands all route through the gateway transparently. 
You: @config.ts Why is the database connection timing out?
Continue: Looking at your config, the connection pool max is set to 5...
The AI saw tokenized credentials. You got an answer that references your actual config. No difference in your workflow.
Continue’s config supports multiple models. You can route some through Takumo and leave others direct. Useful if you want protection on sensitive projects but raw speed on throwaway scripts.

​
TypeScript config

If you use config.ts instead of config.json: 
// ~/.continue/config.ts
import type { ContinueConfig } from "@continuedev/config-types";

export function modifyConfig(config: ContinueConfig): ContinueConfig {
  config.models = [
    {
      title: "Claude via Takumo",
      provider: "anthropic",
      model: "claude-sonnet-4-5-20250929",
      apiBase: "https://gateway.takumo.io/v1",
      apiKey: process.env.TAKUMO_API_KEY,
    },
  ];
  return config;
}
This lets you pull the API key from an environment variable instead of hardcoding it.

​
Troubleshooting

Check the basics:
  1. The gateway URL is correct and reachable: curl https://gateway.takumo.io/v1/health
  2. Your API key is valid (check the dashboard)
  3. The provider field matches what you’re actually using (anthropic, openai, mistral, etc.)
  4. Restart VS Code after config changes
Autocomplete uses a separate config key (tabAutocompleteModel). Make sure you’ve set apiBase there too, not just in models.
Check Supported Patterns for the full list of what Takumo catches. If your secret format isn’t covered, you can add custom patterns in your organization settings on the dashboard.