Skip to main content

​
Generic Setup

Any AI tool that lets you set a custom API endpoint works with Takumo. The pattern is always the same.

​
Three steps

1

Find the endpoint setting

Look for a setting called “API base URL”, “API endpoint”, “Custom endpoint”, or similar in your AI tool’s configuration. Every tool that supports custom providers has one.
2

Change the URL

Replace the provider’s default URL with your Takumo gateway URL:
https://gateway.takumo.io/v1
Self-hosted: use your own gateway URL instead.
3

Set your API key

Use your Takumo API key wherever the tool asks for a provider API key. The gateway handles provider authentication on your behalf.
That’s it. Requests go through the gateway. Secrets get tokenized. Responses get rehydrated.

​
Common provider URLs

Here’s what to replace:
ProviderDefault URLTakumo URL
Anthropichttps://api.anthropic.comhttps://gateway.takumo.io/v1
OpenAIhttps://api.openai.comhttps://gateway.takumo.io/v1
Mistralhttps://api.mistral.aihttps://gateway.takumo.io/v1
Google AIhttps://generativelanguage.googleapis.comhttps://gateway.takumo.io/v1
The gateway auto-detects which provider you’re targeting based on the model name in the request.

​
Environment variable pattern

For CLI tools, the integration is usually two environment variables: 
# Anthropic-based tools
export ANTHROPIC_BASE_URL="https://gateway.takumo.io/v1"
export ANTHROPIC_API_KEY="your-takumo-api-key"

# OpenAI-based tools
export OPENAI_BASE_URL="https://gateway.takumo.io/v1"
export OPENAI_API_KEY="your-takumo-api-key"
Any tool that reads these standard environment variables will route through Takumo without any code changes.

​
Config file pattern

For tools with JSON or YAML configuration: 
{
  "apiBase": "https://gateway.takumo.io/v1",
  "apiKey": "your-takumo-api-key",
  "model": "claude-sonnet-4-5-20250929"
}
The exact field names vary by tool, but the values are always the same: your gateway URL and your Takumo API key.

​
SDK pattern

If you’re building your own integration with an AI SDK: 
import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
  baseURL: "https://gateway.takumo.io/v1",
  apiKey: "your-takumo-api-key",
});

const response = await client.messages.create({
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [{ role: "user", content: codeWithSecrets }],
});
// Secrets in codeWithSecrets were tokenized by the gateway
// Tokens in the response were rehydrated by the gateway

import anthropic

client = anthropic.Anthropic(
    base_url="https://gateway.takumo.io/v1",
    api_key="your-takumo-api-key",
)

response = client.messages.create(
    model="claude-sonnet-4-5-20250929",
    max_tokens=1024,
    messages=[{"role": "user", "content": code_with_secrets}],
)

​
If your tool doesn’t support custom endpoints

Some tools don’t let you change the API URL. For those, use the Takumo CLI to pre-process files before pasting or attaching them: 
# Tokenize a file before pasting into any AI tool
takumo-aegis tokenize src/config.ts
This prints the tokenized version to stdout. Copy and paste it into your AI tool. When you get a response back, rehydrate it: 
# Pipe the AI response through rehydrate
echo "AI response here" | takumo-aegis rehydrate
Or use the shield command for a full round-trip: 
takumo-aegis shield src/config.ts --prompt "Add connection pooling"
If your tool doesn’t support custom endpoints, use the Takumo CLI to pre-process files. Run takumo-aegis tokenize on your code before pasting it into any AI tool, and takumo-aegis rehydrate on the response to restore secrets.

​
Verifying it works

After configuring any tool, verify secrets are being caught:
  1. Open a file with a known secret (database URL, API key, etc.)
  2. Send it to your AI tool
  3. Check the Takumo dashboard — you should see the request logged with detections
  4. If the detection count is zero, check Supported Patterns