Zoo Code

Connect the actively maintained VS Code coding agent to TokenMix.

What Zoo Code Is

Zoo Code is an AI coding agent inside VS Code. It can explain files, generate code, refactor, debug, and work through Code, Architect, Ask, Debug, and Custom Modes. For beginners, the safest workflow is simple: ask questions first, request a plan second, and only then allow file edits.

Zoo Code can connect to TokenMix through the OpenAI Compatible provider. You do not need to deploy Zoo Code or write an adapter. You only need to change the API endpoint in the extension settings.

Prepare TokenMix Values

Prepare these values first:

• Provider: OpenAI Compatible
• Base URL: https://api.tokenmix.ai/v1
• API Key: your TokenMix API key
• Model: a TokenMix model ID, such as <your-model-short-id>

The Base URL must include /v1. Do not use the default OpenAI endpoint, and do not use only https://api.tokenmix.ai.

Install Zoo Code

1. Open VS Code.
2. Open Extensions.
3. Search for Zoo Code.
4. Check that the extension ID is ZooCodeOrganization.zoo-code.
5. Click Install.
6. Open the Zoo Code panel or the Zoo Code settings from the command palette.

If you are migrating from another VS Code coding agent, start with only the model configuration. After chat and code generation work, move MCP, command permissions, and project rules.

Configure TokenMix API

1. Open Settings or Provider settings in Zoo Code.
2. Find API Provider or Model Provider.
3. Select OpenAI Compatible.
4. Set Base URL to https://api.tokenmix.ai/v1.
5. Set API Key to your TokenMix key.
6. Set Model to a TokenMix model ID.
7. Keep advanced options such as Reasoning, Responses API, Embedding, Code Index, and Context Window at their defaults for the first test.
8. Save the settings.

API keys are normally stored in VS Code Secret Storage. Do not put your key in .env, README files, project config files, or screenshots.

First Test

Use a temporary project or a test branch first. Do not let the agent modify a production repository immediately.

Start with Ask mode:

Explain what the current file does. Do not modify any files.

Then try Architect mode:

Create a plan for a classic frontend-backend separated architecture diagram, including frontend, backend, database, cache, and object storage. Do not modify files.

After the answers look normal, try a tiny Code mode task:

Create a small TypeScript function that validates an email string and add one simple usage example.

If these 3 tests work, your TokenMix Base URL, API key, and model ID are configured correctly.

Recommended Order

1. Ask: explain code without editing files.
2. Architect: write a plan before implementation.
3. Code: make a small scoped edit.
4. Debug: investigate one clear error.
5. Custom Modes: standardize repeated team workflows.
6. MCP Servers: connect external tools after the model connection is stable.

This order is safer for new users. An IDE agent can read and write files and run commands, so begin with small tasks and learn the permission prompts before assigning larger work.

MCP

Zoo Code supports MCP Servers. MCP can let the agent call extra tools, such as reading external docs, using design tools, running read-only database queries, or calling internal services.

Recommended setup:

1. Configure only the TokenMix model first.
2. Confirm normal chat works.
3. Add one simple MCP Server.
4. Check that the agent can see the tool list.
5. For the first MCP call, use a read-only task.
6. Confirm permissions separately for file writes, command execution, or database changes.

Do not configure the model and MCP at the same time during the first setup. If something fails, separate steps make it much easier to find the cause.

Code Index And Embedding

If you enable Code Index or repository search, you may also need an embedding model.

• Chat models and embedding models can be configured separately.
• If the UI supports OpenAI Compatible embeddings, you can use TokenMix embedding models.
• Embedding Base URL should also be https://api.tokenmix.ai/v1.
• Model should be a TokenMix embedding model ID.
• If the UI asks for dimension, use the dimension required by the selected embedding model.

Beginners can leave Code Index disabled at first. Normal chat, small edits, and code explanation can work before repository indexing is added.

Troubleshooting

• 401 or unauthorized: the API key is wrong, expired, or has extra spaces.
• model not found: the model is not a TokenMix model ID, or your account cannot access it.
• Base URL error: the Base URL is missing /v1; use https://api.tokenmix.ai/v1.
• Requests still go to OpenAI: the provider is not OpenAI Compatible, or the Base URL is still the default value.
• Output stops early: use a larger-context model or reduce the number of files sent at once.
• Tool call denied: check Zoo Code permission prompts and VS Code workspace trust.
• MCP does not work: confirm normal chat first, then check whether the MCP Server is running and the config path is correct.

Beginner Advice

Treat Zoo Code as an assistant that can write code but still needs your review. Ask it to explain first, plan second, and edit last. Be extra careful with payment, login, database, production config, and secret files. After every change, review the diff and run your project tests or build.