Connect Roo Code to Tokens

Roo Code is a VS Code extension for agentic coding. It connects to Tokens through its OpenAI Compatible provider, which uses the OpenAI Chat Completions protocol against https://tokens.bd/v1.

If you already have Roo Code set up and want to keep it, the steps below work with Tokens.

Configure the OpenAI Compatible provider in Roo Code#

1. Get a key and a model ID#

Create a key in the dashboard; API keys covers spend caps and allow-lists. This guide uses deepseek/deepseek-v4.1-flash. Other IDs are in the model catalog.

2. Enter the provider settings#

Open the Roo Code panel, go to its settings, and set:

These match the Roo Code snippet on Connect Your Agent, which fills in your real key.

Roo Code takes the key in its settings field. It doesn't read TOKENS_API_KEY from the environment.

3. Fill in the model configuration#

Roo Code can't learn model limits from the endpoint, so it asks:

• Context Window. Use the model's context window from the catalog. Roo Code manages conversation length based on this number.
• Max Output Tokens. The model's output limit or less.
• Image Support. On only if the model accepts images.
• Computer Use. Tool use. Leave it on for models that support tool calling.
• Input/Output Price. Only feeds Roo Code's cost display. Tokens bills from its own meter, and your dashboard usage is the real figure.

Tool calling is required#

Per Roo Code's docs, it uses native tool calling exclusively, with no XML fallback. The model you pick must support OpenAI-style function calling. A model without it will answer in chat but fail as soon as Roo Code tries to read a file or run a command. Check the model's capabilities in the catalog before switching.

Switch models#

Change the Model field to another exact Tokens ID and update the context window and output values. List the IDs your key can use with:

curl -s https://tokens.bd/v1/models -H "Authorization: Bearer $TOKENS_API_KEY"

Choosing a model explains how to pick one for agent work.

Verify it works#

Test the key and model directly:

export TOKENS_API_KEY=tok_live_your_key
curl -s https://tokens.bd/v1/chat/completions \
  -H "Authorization: Bearer $TOKENS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "deepseek/deepseek-v4.1-flash", "max_tokens": 16, "messages": [{"role": "user", "content": "Reply with OK"}]}'

Then ask Roo Code to do something that needs a tool, such as "Read package.json and tell me the project name." A successful tool call confirms both the connection and function calling. The request should appear in your dashboard usage analytics.

Troubleshooting#

Errors about tools or tool calls. The model doesn't support native tool calling, or handles it poorly. Pick another model with tool support.

401 invalid_api_key. Wrong, revoked or rotated key. Rotation stops the old secret immediately.

404 model_not_found. The model ID needs the full provider prefix, for example deepseek/deepseek-v4.1-flash.

404 on every request. The base URL must be exactly https://tokens.bd/v1.

403 model_not_allowed_on_key. The key's allow-list excludes the model. Allow-lists are fixed at creation, so create a new key.

429 rate_limited, concurrency_limit or window_exhausted. Per-minute limit, concurrent request limit, or plan usage window. Each response includes Retry-After.

Something broke after a VS Code update. Roo Code won't receive a fix. Move to Kilo Code or Cline; the Tokens settings carry over almost one to one.

See Troubleshooting for every error code and Chat Completions for the request format.

Sources: Roo Code docs, OpenAI Compatible provider (last updated 2026-05-15), RooCodeInc/Roo-Code repository (archived), checked October 2026.