Complete settings guide

Settings controls how Nexts behaves and where configuration and data are stored on this computer. Select the gear at the bottom of the activity bar, then choose a category in the Settings sidebar.

Before changing settings

• Most display and interaction preferences apply immediately. Service, directory, and network changes may require a refresh, reconnection, or restart.
• API keys, access tokens, pairing QR codes, and local paths may be sensitive. Redact them before taking screenshots or recordings.
• When unsure, change one option at a time and test it. This makes troubleshooting and rollback easier.
• Options can vary by client version and operating system. Follow the labels and controls shown in your installed client.

General

General settings collect everyday preferences such as language, startup behavior, conversation interaction, and notifications.

1. Show a system notification for unread conversation updates.
2. Show a system notification when a scheduled task runs.
3. Show a floating orb in the screen corner for quick access to features like audio recording & transcription.

If notifications do not appear, also check whether Windows allows notifications from Nexts. See Tasks and schedules for scheduled runs and result history.

Display

Display settings affect presentation only; they do not alter conversations, projects, or model data. Depending on the installed version, they can include theme, interface scale, Markdown and code rendering, or layout preferences.

After changing scale or layout, check a long conversation, code block, table, and sidebar. If larger text pushes controls off-screen, restore the default scale and increase it one step at a time.

1. Show the sender's avatar and name above each message; hidden by default.
2. Document mode: all messages left‑aligned.Chat mode: my messages on the right, AI messages on the left.
3. Show the model context‑window usage chip in the chat composer.
4. Show a microphone button in the composer for push‑to‑talk dictation (off by default).

Agents

The Agents page lets you select the local agent runtime to power your chat. Nexts AI is the built‑in system‑default agent. Once you have installed Claude Code or Codex CLI locally on your machine, they will be auto‑discovered via ACP bridge and appear in this list. Click the Refresh button after installing new external agents to rescan your system. Note: Agent runtime and model services are separate configurations.

Model Services

1. Nexts AI (Cloud / Official): Built‑in official provider with official quota. Toggle the switch to enable cloud models; no extra API key is required.
2. Ollama (Local): Local model connector. You need to install and launch the Ollama application on your computer beforehand. No API key is needed for local Ollama service. After Ollama is running, click Fetch model list to load your locally downloaded models. If no local models are detected, verify that Ollama is up and running.
3. Click Add Provider to add third‑party AI model providers.

If no model appears, fetch model list, check the network, service URL, key validity, account access, and enabled state. Never paste a complete API key into screenshots, logs, or feedback.

Media Services

1. Microphone Device: Select the microphone used for recording and meeting transcription. Choose “System default” to follow the operating system’s current default input device, or select a specific connected microphone.
2. System Audio Device: Select the system output device to capture through loopback recording. This is used to record the far-end audio in a meeting. Normally, choose the speaker or headset playing the meeting audio, or use “System default.”
3. Transcription Engine: Choose the speech recognition engine used to convert recordings into text. Nexts supports local Whisper, local Parakeet, configured remote audio models, and the Nexts Official cloud service. Whisper is robust and general-purpose; Parakeet is faster but requires an approximately 670 MB download and does not support Chinese. Remote models must first be configured under Model Services, while the official cloud service requires a Nexts account.
4. Transcription Language: Specify the language spoken in the recording. Available options currently include Auto-detect, English, Italiano, and Simplified Chinese. Pinning a language can improve recognition stability and prevent live transcription from mistaking Chinese for English. Do not select Chinese when using Parakeet.
5. Model Folder: Shows where local speech transcription models are stored. Click “Open folder” to view or manage local model files, such as Whisper .bin models. This folder is primarily used by local transcription engines and is not required for cloud transcription.
6. Summary Model: Manage the local large language model used to turn meeting transcripts into structured meeting minutes. The model runs through Ollama, and the recommended default is typically qwen2.5:7b. Ollama must be installed and running before the model can be downloaded. The same button can also be used to download it again.
7. Summary Model Source: Choose the model that actually generates meeting minutes. “Default (follow current selection)” uses the current model selection, or you can explicitly select an enabled text-capable local or cloud model configured under Model Services.
8. Start a Real-Time Conversation: Click “Open voice orb” to start a real-time voice conversation with an expert. You can interrupt while the expert is speaking. The real-time voice provider is determined by the Nexts server configuration. This feature is currently available on desktop and requires signing in to a Nexts account.
9. Voice Wake: When enabled, Nexts listens for the configured wake phrase on desktop. Wake-phrase detection is performed locally with a Whisper model, and the detection audio is not uploaded. Once the phrase is recognized, dictation starts in the current chat composer. The subsequent command dictation uses the official cloud voice service and therefore requires signing in. If the current desktop build does not include local Whisper support, this feature is shown as unavailable.
10. Wake Phrase: Set the phrase that starts voice dictation, such as “Hey, Nexts.” The wake phrase must contain between 2 and 40 characters. Press Enter or click outside the field to save it. This setting takes effect only when desktop voice wake is available.
11. Auto-Play AI Replies: When enabled, Nexts automatically synthesizes and plays each expert reply after the complete response has been generated. Turning this option off stops the current voice playback and prevents future replies from being read aloud automatically.
12. Voice: Choose the voice used when AI replies are read aloud automatically. Available voices are provided by the Nexts Official voice service and are loaded after signing in. This setting affects only speech playback; it does not change the written response or the chat model being used.

See Nexts Canvas for image and video workflows. Media inputs may be sent to the selected provider, so review that provider's data handling rules first.

Experts

Experts are reusable assistants with a defined role, tools, and skills. Use these settings to maintain the default model, instructions, available tools, and memory scope.

Define the task and stop conditions first, then grant only the required tools. Test one representative task and one boundary case to confirm both usefulness and scope. See Workflows for using experts in a workflow.

Memory

Memory lets future conversations reuse durable preferences and project context. User memory is appropriate for stable personal preferences; project memory should contain conventions for one project.

• Save only information that is useful over time and safe to reuse. Never store passwords, tokens, identity documents, or one-time sensitive content.
• If an answer uses stale information, inspect, update, or delete the relevant memory before trying again.
• Turning memory off may only stop future capture or use. Use any separate delete control to remove existing entries.

1. Long-Term Memory:Controls the long-term memory feature in Nexts. When enabled, AI experts can read and use stored user and project memories in future conversations, allowing responses to better reflect your preferences and the current project context. Disabling this feature does not delete existing memories; it pauses their use and related automatic processing.
2. Automatic Memory:When enabled, Nexts automatically identifies and saves durable information after each completed expert turn, such as user preferences, feedback, project constraints, and important references. Automatically extracted memories are saved without individual review. Temporary task state, secrets, and private file contents should not be stored as long-term memory.
3. Reflect with Model:Uses the active chat model to analyze completed conversations and extract higher-quality, better-structured long-term memories. The model determines which information is worth retaining and whether it belongs to user or project memory. If model-based reflection fails or no model is available, Nexts automatically falls back to its built-in rule-based extractor.
4. AutoDream:Periodically organizes long-term memory in the background by consolidating duplicate information, removing stale or excess entries, and reducing redundancy. A consolidation run occurs only after both the configured time interval and minimum number of completed turns have been reached. Settings:
   • Hours: The minimum time between AutoDream runs, from 1 to 168 hours.
   • Turns: The minimum number of completed expert turns required before another run, from 1 to 50.
   • Max entries: The maximum number of entries retained in each memory scope, from 20 to 500. For example, with 24 hours and 5 turns configured, AutoDream runs again only after at least 24 hours have passed and at least five additional turns have been completed.
5. Add Memory: Manually add information that the AI should remember over time. First select the memory scope, then enter a key and value. The key identifies the type or topic of the memory, such as writing_style, while the value contains the actual information, such as “Keep responses concise and direct.”
   • User memory: Applies to every conversation and is suitable for language preferences, writing style, and long-term working habits.
   • Project memory: Applies only to the current project or workspace and is suitable for project conventions, technical constraints, and business context. A project must be open before project memory can be added. Saved memories appear in the list below and can be edited or deleted at any time.

Security Center

Security Center controls authorization boundaries for local capabilities, tool calls, and sensitive actions. See Permissions and safety for recommended policies.

Remote Access

Remote Access manages the local remote service, LAN access, public tunnel, and mobile pairing. Use the phone icon in the title bar for routine pairing; open Settings to inspect state, change network scope, or troubleshoot. See Mobile and remote access.

Pairing QR codes and copied links contain short-lived credentials. Share them only with trusted people and devices. Enable LAN access only on a trusted network, and stop the tunnel or service when no longer needed.

1. Enable Remote Service:Starts the local Nexts Edge remote access service. Once the service is running, a phone or browser can connect to Nexts on this computer through the displayed access URL. Enabling this option alone primarily provides local access. To connect from another device on the same network, enable “Allow LAN / remote access.” To connect from outside the local network, enable “Public tunnel.” Stopping the remote service also stops the active public tunnel.
2. Allow LAN / Remote Access:Allows Nexts Edge to bind to the computer’s LAN address so that phones, tablets, and other computers on the same network can access Nexts. The remote service automatically restarts when this setting is changed so the new network configuration can take effect. Use this feature only on trusted networks, such as your home or office network. Once enabled, other devices on the same LAN may be able to discover and access the service. Do not enable it on untrusted networks such as public Wi-Fi in airports or hotels.
3. Public Tunnel:Creates a public access URL through the managed Nexts relay, allowing a phone to connect to Nexts on this computer even when it is outside the computer’s local network. Manual router port forwarding is not required. The public tunnel requires a signed-in, online, and device-bound Nexts desktop client. Once connected, the page displays a “Public online” status and a remote.nexts.ai access URL, together with a QR code for mobile pairing. Scan the QR code with the Nexts mobile app to connect, or copy the pairing link to your phone. The pairing token in the QR code is time-limited. The page shows its remaining validity and refreshes it automatically, or you can select “Refresh” to generate a new pairing code immediately.
4. Mobile Connection Shortcut:The phone icon in the title bar provides quick access to mobile pairing. Clicking it opens the QR-code panel. If the local remote service or public tunnel is not running, Nexts attempts to enable them automatically before generating a mobile pairing QR code. A green dot on the phone icon indicates that the public tunnel is currently active. The QR code in the quick panel is time-limited and is renewed automatically before it expires. You can also copy the pairing link or refresh the QR code manually. This shortcut requires the desktop client to be online and signed in to a Nexts account. Opening the panel may automatically start the remote service and public tunnel.

Data Settings

Data Settings manages local storage locations, space usage, and conversation cleanup.

Change the data or project directory

1. Finish running conversations, tasks, and media generation.
2. Use the directory-change control and choose a durable location with read and write access.
3. Let Nexts finish migration or reload. Do not exit or move source files during this step.
4. Select Open and verify the new path and existing content.
5. Create a test conversation or project and confirm it is written before cleaning up the old directory.

Avoid temporary folders. For cloud-synced or network drives, first confirm continuous availability and database file-locking support; use a local disk when uncertain.

Use logs for troubleshooting

Reproduce the problem once and note the time, then inspect logs from the same period. Share only relevant excerpts and remove account details, paths, tokens, and third-party response content.

Clear all conversations

Clear all conversations permanently deletes every conversation and cannot be undone. Export or back up anything you need first, and make sure no conversation or task is running. To hide history without deleting it, use Archive Chats.

Keyboard Shortcuts

1. Open Settings → Keyboard Shortcuts.
2. Select a command, press the new key combination, and save it.
3. Test global shortcuts with Nexts focused and from another application.

If a shortcut does not respond, it may conflict with Windows, an input method, or another application. Avoid overriding system security or accessibility shortcuts. See Using Nexts.

Archive Chats

Archive Chats removes conversations from the everyday history list while retaining them for later viewing and restoration. It is different from deletion. Archive completed long conversations to keep history focused.

About

About shows the current Nexts version, update information, and related links. When reporting a problem, include the app and OS versions, occurrence time, and reproduction steps—never passwords, API keys, active QR codes, or unredacted private content.

If the interface differs after an update, confirm the client version, then check release notes or Troubleshooting.

Where account settings live

Sign-in, registration, subscriptions, devices, and avatar management live on a separate account page. Select your avatar or the Nexts logo at the top of the activity bar to open it.