Agents, models, and usage — Noea Code documentation

The Agent reads your project and performs tasks; the model interprets and generates content. This guide covers managing Agents, connecting model providers, choosing settings, and reading usage information. Available models, sign-in methods, and features depend on the selected Agent.

Manage local Agents

Open Settings → Agent. Noea currently offers Codex, Claude Code, OpenCode, and CodeBuddy Code, each with its version, status, and main action.

Click an Agent's row to open its details panel for installation, sign-in, checks, updates, or uninstalling. Each Agent is prepared separately. Closing the panel does not interrupt preparation already underway.

Ready means verification passed. An installed status or detected credentials still require verification. For failures, follow the details panel. See Install and get ready for initial setup.

Connect a model provider

Agents that support multiple providers show a provider list in their details panel. Select a provider and connect using an account or API key as instructed.

The same provider is connected and managed separately in each Agent. The Agent handles sign-in and credential storage; Noea does not separately store sign-in keys. The Agent may still store authentication data locally.

OpenCode also offers a custom endpoint option. Enter a name, Base URL, protocol, and at least one model, then save and configure credentials in the sign-in window. Choose Chat Completions or Responses to match your service's API.

A failed provider-list read is different from an unconnected provider. Refresh to check the current connection state and retrieval time.

Configure the current chat

Use the settings menu below the input to choose the Agent, model, and reasoning effort. Reasoning effort adjusts how much reasoning the model devotes to a problem. Models are grouped by provider. Reasoning effort, plan mode, fast mode, and other options appear according to Agent support.

A new chat normally uses the Agent from the most recent successfully connected regular chat. You can switch before the first message, but not after connection. Start a new chat to use another Agent.

Some model lists are only available after connecting, which the interface indicates. If the Agent does not accept an option, it will not take effect and the chat explains why.

Permissions have a separate control in the input. Choose from the Agent's available options according to the task.

Configure project operations

In Settings → Current Project → Operations, choose an Agent, model, and reasoning effort separately for Realize, Explain, and Organize Source.

These settings apply to their respective project operations, not all regular chats. Explain triggered after implementation in a chat uses that chat's Agent. Explain following an accepted organization plan uses the project's Explain settings.

If the selected Agent is unavailable, Noea uses an available alternative and shows a notice. Check the effective choice here before running an operation.

Read the current chat's usage

The small ring below the input shows context usage. Click it to view context, account limits or balances, and token and cost information for the current chat. Tokens are units used by models to process text and other content; they are not the same as words or characters.

Context usage indicates how much of the model's available context space the chat uses. Account limits describe usage restrictions imposed by the provider or Agent. They measure different things.

View totals and interpret costs

Open Settings → Usage, or choose the option to view all usage in the usage panel. Filter by today, the last 7 days, the last 30 days, or all time. Expand an Agent for model and token details.

Information	Scope and meaning
In-app usage	Records generated by work within Noea.
Account limits	May include usage outside Noea and do not change with the period filter.
API list-price equivalents or estimates	A way to understand usage scale, not your actual subscription bill.
Credits	Shown separately by source, not treated as US dollars.
Unpriced or unknown	Missing prices or readings, not zero cost or zero usage.

Readings show their update time. A failed refresh retains the previous value. Refresh after a reset time before interpreting the limit. If an Agent cannot report account limits, the interface states that they are unavailable.

For connection issues, see Troubleshooting. For extensions, see Add MCP servers and Skills.