Models & Providers
This page covers the part of /agent-settings where you decide whether the built-in steward can run and which models it can use.
Enable AI agent
The Enable AI agent toggle turns the built-in Web UI assistant on or off.
You can also enable it through configuration:
export DAGU_AGENT_ENABLED=trueModels Table
When AI agent is enabled, Dagu shows a models table where you can:
- add a model
- edit an existing model
- set the default model
- delete a model
The table helps you compare:
- display name
- provider
- model ID
- whether credentials are ready
- which model is the default
Default Model
The default model is the one the Web UI steward uses unless you override it in a chat session.
Workflow steps with action: agent.run also fall back to this default unless the step sets its own model.
Add Or Edit A Model
The form includes these common fields:
| Field | What It Means |
|---|---|
Name | Human-friendly label shown in the UI |
ID | Stable short name for this saved model, used in the UI, config, and API |
Provider | Which model service Dagu should call |
Model | The provider's actual model name or tag |
API Key | Credential for hosted providers; optional for Local unless your proxy requires it |
Base URL | Optional endpoint override |
Description | Optional note shown in the UI |
Context Window | Optional capacity metadata |
Max Output Tokens | Optional limit metadata |
Input Cost / 1M tokens | Optional pricing metadata |
Output Cost / 1M tokens | Optional pricing metadata |
Supports Thinking | Whether this model should expose reasoning controls in Dagu |
Reasoning Effort | Default reasoning depth when thinking is enabled |
On creation, Dagu can generate the ID from the Name. After a model exists, the ID is treated as stable.
Providers In The UI
The provider list may include:
AnthropicOpenAIOpenAI CodexGoogle GeminiOpenRouterOpenCodeLocalZ.AI
Available providers can expand as Dagu adds support for more backends.
Credential Patterns
Most providers use the API Key field.
The main exception is OpenAI Codex, which uses a ChatGPT Plus/Pro login flow instead of a raw API key. For that path, see OpenAI Subscription.
Local Models
For Ollama and other local servers:
- choose Provider =
Local - set Model to the exact local model tag
- leave Base URL empty or set
http://localhost:11434/v1
See Local AI for setup and troubleshooting.
OpenCode
For Kimi, DeepSeek, GLM, Qwen, and other models via an OpenCode subscription:
- choose Provider =
OpenCode - set Model to the model ID (e.g.
kimi-k2.6,deepseek-v4-pro,glm-5.1) - leave Base URL empty (defaults to
https://opencode.ai/zen/go/v1) - set API Key to your OpenCode API key
Models that support reasoning (Kimi K2, DeepSeek V4 Pro, GLM-5, Qwen3) work with Supports Thinking enabled.
See OpenCode for available models and workflow YAML examples.
A Practical Rule
Not every field matters equally for every provider. Focus on the fields your provider actually needs:
- hosted models usually need provider, model, and API key
- local models usually need provider, model, and sometimes base URL
- pricing and context fields mostly help Dagu label and reason about the model in the UI