DeepSeek + TOKNADO

How to use DeepSeek models in your AI tools through one endpoint

DeepSeek's models are popular for coding and agent work, and DeepSeek's own API follows the OpenAI chat completions format. That means any tool with a custom base URL field can use them, and so can any OpenAI-compatible gateway that serves them.

This page covers the part that trips people up: choosing the right model ID, testing with curl and Python, and wiring the result into the coding tools you already use. Availability of specific DeepSeek models through TOKNADO depends on what the managed pool serves, so confirm the model IDs in your dashboard.

Start free trial Jump to setup

What you need

• A TOKNADO endpoint and key from your dashboard.
• The exact model ID for the DeepSeek model you want, as shown in the dashboard.
• Any OpenAI-compatible tool, or just curl and Python for testing.

Setup: DeepSeek with TOKNADO

1. Test with curl first

   Prove the endpoint, key and model work before configuring any tool. If this fails, no tool will work.

   curl https://toknado.up.railway.app/v1/chat/completions \
     -H "Authorization: Bearer YOUR_TOKNADO_API_KEY" \
     -H "Content-Type: application/json" \
     -d '{
       "model": "toknado/auto",
       "messages": [{"role": "user", "content": "Reply with the word ready."}]
     }'

2. Test with the OpenAI Python SDK

   The official SDK works by changing the base URL and key. Nothing else in your code changes.

   from openai import OpenAI
   
   client = OpenAI(
       base_url="https://toknado.up.railway.app/v1",
       api_key="YOUR_TOKNADO_API_KEY",
   )
   
   reply = client.chat.completions.create(
       model="toknado/auto",
       messages=[{"role": "user", "content": "Write a Python function that reverses a string."}],
   )
   print(reply.choices[0].message.content)

3. Wire it into your coding tool

   In any tool with an OpenAI-compatible option, set the same three values: base URL, API key, model ID. Setup guides: Cline, Roo Code, Aider, OpenCode, Open WebUI.

4. Pick the model for the job

   DeepSeek has offered a faster, cheaper tier for high-volume loops and a stronger tier for harder reasoning and review. Use the faster one for autocomplete-style and sub-agent work, and the stronger one when output quality matters more than speed. Check DeepSeek's current docs and your TOKNADO dashboard for the live model names.

Notes for real-world use

• Reasoning models can produce long outputs. Set sensible max-token limits in your tool so one request doesn't run away.
• DeepSeek's API supports function calling in the OpenAI format, which is what terminal agents rely on for reading and writing files. Whether tool calls work through any gateway is worth a quick test before a long session.
• Each tool call round trip is a request. Agent loops with a fast model can use many requests per task, so compare against your plan's rolling hourly limit.

Troubleshooting DeepSeek + TOKNADO

FAQ