Connect OmniDino to TRAE, power AI coding with your own API key.
TRAE is a desktop AI IDE. By adding a custom OpenAI model, you can connect OmniDino’s Chat Completions endpoint and use your models and balance across Chat, Agent, and project workflows.
What Is TRAE?
It is a desktop IDE built around AI coding, suitable for everything from code completion to multi-step agent development tasks.
The editor and agent share the same workspace.
TRAE can import existing projects, read code context, explain code in chat, and let agents edit files or run commands. When using a custom OmniDino model, first verify tool calling and code quality in a test Git branch.
bytedance/trae-agent, a Python-based general software engineering agent that supports custom Base URLs through YAML configuration, including OmniDino.
Official client and open-source GitHub tools.
TRAE IDE
A desktop AI IDE. The client itself is not fully open source, so download its installer from the official website.
TRAE Agent
ByteDance’s open-source software engineering agent supports OpenAI, Anthropic, Gemini, OpenRouter, Ollama, and custom Base URLs.
Git
Used for version control, branches, diffs, and rollbacks. Installing it before letting an agent edit code is strongly recommended.
Visual Studio Code
An open-source code editor that can serve as a backup editor and help compare TRAE-edited files and extension compatibility.
Node.js
An open-source runtime required by some MCP servers, front-end projects, and npm tools.
MCP SDK and Reference Servers
Use the official MCP SDK and reference servers when developing tools for TRAE Agent or custom agents.
Install TRAE and import your first project.
Windows
Download the installer from the official TRAE website, then import a local Git project after installation.
src
app.js
api.js
public
README.md
package.json
macOS
Download the build for your architecture, drag TRAE into Applications, and allow macOS to open it.
components
Header.js
Footer.js
styles
README.md
package.json
Add a custom OpenAI model in four steps.
TRAE setting labels may vary by version, but the core fields remain the same.
Open Settings and Models
From the TRAE main window, open Settings in the upper-right corner and go to Models, AI Services, or the model management section.
Click “Add Model”
Choose to add a custom model and set the provider to OpenAI.
Enter OmniDino Details
Copy the model ID exactly from the OmniDino model marketplace, use your OmniDino API key, and enter the complete Chat Completions URL.
Save and Switch Models in Chat
After saving, return to Chat or Agent, manually select the newly added model, and send a minimal test message.
OmniDino Configuration
Complete each field in the TRAE custom model dialog.
https://api.omnidino.com/v1.
Provider: OpenAI
Model: Custom Model
Model ID: Enter the exact OmniDino model name
API Key: Paste your OmniDino API key
Custom Request URL:
https://api.omnidino.com/v1/chat/completionsVerify the connection before allowing the agent to edit code.
Reply with only:
OmniDino TRAE connection successful.
Do not read files, run commands, or modify the project.Check the Response
The model should respond normally without authentication errors, 404 errors, or model-not-found messages.
Check the Active Model
Confirm that Chat is using your custom model rather than the default TRAE model.
Check OmniDino Usage
Open the console and confirm the request, model name, and usage charge are recorded.
How Should Each Mode Be Used?
Chat Mode
Best for explaining code, generating snippets, analyzing logs, and discussing solutions, with the lowest risk.
Agent Mode
Best for multi-file tasks and command execution. Always use a Git branch and review the diff.
SOLO Mode
Best for more autonomous product and engineering workflows, with AI Services and tool integrations.
MCP Tools
Add databases, browsers, GitHub, or internal business tools to custom agents.
Max Mode
Useful for longer context and complex tool calls, but it usually increases model usage.
Project Rules
Define the technology stack, prohibited actions, test commands, and output format for the agent.
TRAE configuration compatibility reference.
| Scenario | Recommended Configuration | Notes |
|---|---|---|
| OmniDino Chat Models | /v1/chat/completions | Use an OpenAI custom model and the complete request URL. |
| OmniDino Responses Models | Not recommended for this configuration | The TRAE custom model page is primarily designed for Chat Completions. |
| Native Claude Messages | Depends on available TRAE provider options | Do not enter a Messages endpoint in an OpenAI Chat Completions configuration. |
| Image Generation | Not a standard chat model | /v1/images/generations cannot be used as a chat model endpoint. |
| Model ID | Copy the exact model name | Do not rename it or enter a display label. |
| API Key | Dedicated TRAE Key | Set a separate quota to simplify risk control and usage tracking. |
Use TRAE Agent when you need a fully open-source option.
TRAE Agent is a standalone CLI project that does not require TRAE IDE and supports custom Base URLs.
TRAE Agent + OmniDino
The official repository supports YAML configuration and a custom base_url for the OpenAI provider.
model_providers:
omnidino:
api_key: "Paste your OmniDino API key"
provider: openai
base_url: "https://api.omnidino.com/v1"
models:
omnidino_model:
model_provider: omnidino
model: "Enter the exact model name"
max_tokens: 4096
temperature: 0.3
agents:
trae_agent:
model: omnidino_model
max_steps: 80
tools:
- bash
- str_replace_based_edit_tool
- sequentialthinking
- task_doneTroubleshoot connection failures step by step.
Still unable to connect TRAE to OmniDino?
Provide the TRAE version, operating system, model ID, custom request URL, HTTP status code, and complete error message. Hide your API key and never send a screenshot containing the full key.
[email protected]