Skip to content

Settings

AI:GO groups the settings tabs by how often people change them:

  • General: what most people change, such as the language, the theme, models, and providers.
  • Features: the settings of each feature, such as memory, Data, Glance, and Squad.
  • Admin & development: engines, the API server, the organization, and the other settings for administrators and developers.

Every group stays open. A group whose tabs are all hidden by an organization policy is not shown, and on a narrow window the tabs run as one row across the top, without the group headers.

Older docs sometimes referred to retired tabs such as Inference, Network, or Storage. Those names no longer match the current UI.

Settings page with its grouped navigation Settings page in dark mode

Current settings tabs

Group Tab What it covers
General General app language, startup behavior, the chat's temperature and reasoning effort, the system tray and what closing the window does, notifications, updates, privacy, and the app's information and legal documents
General Appearance theme, sidebar, motion, and UI scale
General Models the models directory, additional model folders, default models for app surfaces, and Running on this computer: the GPU layers, the context length, automatic model management, and speed measurement
General Providers cloud model providers and the models used from each
General Voice Input microphone and transcription behavior
General Ask about a window asking about a window on the screen (the desktop app on macOS and Windows)
Features Memory memory-related controls, with the auto-extraction timings under Tuning
Features Data Data Hub search presets and maintenance: watched folders, re-fetch, wiki updates, the data doctor, and the trash
Features Glance what Glance reads, the names it looks for, and per-app settings
Features Squad the messenger channels connected to squads, the agent runner image, the global instructions for squad agents, and how much run history each squad keeps
Features Tools & extensions tool permissions, tool approval rules, MCP, and the web search tool
Features Connectors connector settings
Features Nodes Share this computer and Connect other computers, with load balancing and running across a cluster
Features App Plugins installing and managing app plugins
Admin & development Engines installed inference engines and the ones you can install, default engines by format, and the MLX and container engine settings
Admin & development API Server the API server and its Management API, routing, health, security, keys, Claude Code, mesh, and Agent Protocol (ACP) sections
Admin & development Organization organization-wide settings
Admin & development Advanced logging, ports and networking, Supervisor history, Demo Mode, CLI tools, the system monitor, connection addresses and the network certificate, backup and restore, and reset

Where key controls live now

Downloaded models and running models

Settings has no model list. The Manage models row at the top of Settings > Models opens the Models window on Manage, which lists the models on this computer, cloud models, and network node models, and runs and stops them. A connected node's detail in Settings > Nodes has a row that opens the same view.

Models directory

Open Settings > Models.

Additional model folders

Open Settings > Models > Additional model folders. Besides its own models directory, the app can read models from folders that belong to another application, such as an LM Studio library, the Hugging Face cache, or the Ollama model store. Add a folder with Add folder, or press Find models from other apps to have the app look in LM Studio's default locations (~/.lmstudio/models and ~/.cache/lm-studio/models), in the Hugging Face cache (see Hugging Face cache), and in Ollama's (see Ollama models), and add one with a click.

A model from an added folder is listed and loaded like any other, and the Models page marks it with the folder's name. The folder itself is treated as read-only: the app never writes into it, so Delete is not offered for these models and nothing you do here changes what the other application sees. Settings you make for such a model, including an alias and its launch configuration, are kept in the app's own data directory and survive removing and re-adding the folder.

Each folder can be turned off without removing it, which hides its models until you turn it back on. A folder that is missing, on a drive that is not connected, or that the app cannot read stays in the list as Not available. Its models are hidden while it is unavailable, Refresh on the Models page keeps them in the index, and they reappear when the folder is readable again. Removing a folder and adding it again under the same key restores the settings saved for its models.

A folder cannot overlap the models directory, another added folder, or the app's own data directory, and it cannot be the filesystem root, your home directory, or a folder that contains your home directory (such as /home). At most 16 folders can be added. A symbolic link from the models directory into an added folder is not followed, so the same files are never listed twice and the scan never writes into the folder through the link.

On a server, a folder declared in config.toml takes precedence: if a folder saved in the settings uses the same key, the saved entry is dropped and a warning is logged. See Additional Model Sources.

Ollama models

Models pulled with ollama pull are read where Ollama keeps them, without copying or converting anything. Find models from other apps looks for Ollama's store in the folder OLLAMA_MODELS names, in ~/.ollama/models (%USERPROFILE%\.ollama\models on Windows), and on Linux in /usr/share/ollama/.ollama/models, where the Ollama service install keeps it. When you add a folder by hand, pick that models folder; picking the .ollama folder above it adds the models folder inside it. The app reads Ollama's manifests and loads each model in llama-server straight from Ollama's blob store. Each tag is listed under the name Ollama shows, such as llama3.2:3b or hf.co/Qwen/Qwen3-0.6B-GGUF:Q8_0, so two tags of the same weights are two entries, and they share an alias and launch settings because those are kept per weights file. A vision model whose download includes a projector loads with it and is marked as able to read images.

A model is not listed while its download is incomplete or its weights are missing, when Ollama stores it only as safetensors weights for its MLX engine, or when it is a fine-tuned adapter (LoRA) over a base model. The chat template comes from the model file itself: Ollama's own template is written for Ollama's Go template engine and is not used, so a model whose file has no chat template behaves as it would from the models directory.

The app skips anything in the store that is not a regular file, such as a named pipe or device file where a model file belongs, instead of waiting on it. It reads at most eight entries of a manifest that lists several runners, so a manifest padded with entries is not read in full. A load request over the REST API (POST /models/load) that names a projector (mmprojPath) must name a file inside the models directory or an additional model folder, and any other path is refused with a 400 error. A projector file that has since been deleted, for example one Ollama removed, is ignored and the model loads without it.

Ollama manages its store on its own terms. It deletes a model's files on ollama rm and removes files no model refers to when it starts (unless OLLAMA_NOPRUNE is set), and a model removed there disappears here at the next scan. On Windows, a model loaded here keeps its file open, so Ollama cannot delete it until you unload the model. Some models in the Ollama library use architectures that Ollama's engine runs and llama.cpp does not support yet; they are listed, and loading one fails with the engine's load error.

Hugging Face cache

Models that huggingface-cli download, the huggingface_hub library, mlx-lm, or llama.cpp -hf fetched are kept in the Hugging Face cache, by default ~/.cache/huggingface/hub (%USERPROFILE%\.cache\huggingface\hub on Windows). When HF_HUB_CACHE is set, the cache is that folder instead, and when HF_HOME is set, it is the hub folder inside it. Find models from other apps checks all three. To add the cache by hand, pick the hub folder itself, the one that holds the models--<org>--<name> folders; the app refuses a folder that holds none of them unless it is empty.

The app reads one revision of each model repository: the one named in the repository's refs/main, or the most recently modified snapshot when that file is missing. GGUF files, split GGUF files, and MLX safetensors repositories are listed under the repository's own name (ggml-org/Qwen3-0.6B-GGUF), marked Hugging Face cache, and load from the cache without a copy. A download that is still running or was interrupted is skipped, and datasets and Spaces in the same cache are ignored. As with any added folder, the app never writes into the cache, so huggingface-cli keeps managing it.

The app takes refs/main and the snapshots folder only when the cache itself wrote them as plain entries. A refs folder, a refs/main file, or a snapshots folder that is a link is not followed, and a refs/main larger than a revision hash is not read; in those cases the app uses the most recently modified snapshot, and a linked snapshots folder means the repository is not listed. If you added the cache through a link to the hub folder, its models keep the names they have in the list when you load them. A GGUF file whose header nests data absurdly deep is still listed, but the app does not read its details, and one such file does not stop the scan. Diffusion models in the cache find their companion files (VAE, text encoders) in the same snapshot folder.

Default models for app surfaces

Open Settings > Models.

Default engines by format

Open Settings > Engines and select Advanced.

Default reasoning effort and generation defaults

Default temperature and Default reasoning effort are in the Chat section of Settings > General. The context length a local model loads with is in Settings > Models > Running on this computer. Both were the Generation tab, and an organization policy that hid that tab still hides them.

When Use model's max context is enabled, AI:GO starts with the model's reported native maximum and limits it with Automatic context limit. It can reduce that value further when the launch memory estimate does not fit the available RAM and VRAM. This estimate does not guarantee that a load will succeed. If system metrics are unavailable, or the model estimate does not fit even at 512 tokens, the automatic context limit remains the selected value and the normal compatibility check remains authoritative.

An explicit context chosen for a load has first priority. A saved per-model context has second priority, and both remain exact. Otherwise, AI:GO uses the limited native maximum, then the selected preset, then the custom default.

Default temperature applies to every chat, with local and cloud models alike. A temperature you change for one conversation in Conversation parameters comes first, then a temperature saved in the model's own settings, then this default. A reasoning model asked for a reasoning effort receives no temperature. The maximum response length is set per conversation in Conversation parameters.

MLX engine defaults

Open Settings > Engines.

Running across a cluster

Open Settings > Nodes > Connect other computers. Run across cluster, at the end, holds Enable running across a cluster and Default transport backend. The transport port range stays with the other ports, below.

Run across cluster transport port range

Open Settings > Advanced > Ports & networking.

Socket directory

Open Settings > Advanced > Ports & networking. Leave Socket directory empty to use the per-user runtime directory described in Local Socket Location. A custom value must be an absolute path inside your home directory, 62 bytes or shorter (so a socket name still fits in the 103-byte limit), and is applied to the router, the model servers, the Management API, and the monitor at the next launch. A longer path is rejected when you save, with the byte count in the message.

Running on this computer

Open Settings > Models and select Running on this computer. The row shows this computer and the fastest local model measured here. The detail opens with this computer on a card (its chip or GPU, memory, and accelerator) and, beside it, that model with its answer speed. Below the card are the GPU layers and the context length a local model loads with, Automatic model management (see Supervisor Agent), and Speed measurement: Tune settings automatically, the check on a model's first run (off by default, so nothing is measured unless you ask), and how long records are kept (90 days by default; baselines stay).

Benchmark, at the end, opens the Benchmark page's Measure tab, which lists each local model with its answer speed and first response on this computer, measured by a 15-second Quick check. A model's detail, which model cards, notifications, and statistics open, sits one level further in. It adds the verdict for chat and the engine and settings the model was measured with, and This model's records opens the model's history on the Benchmark page, where comparing runs and raw llama-bench measurement live.

This was the Performance tab. An organization policy that hid that tab or the /benchmark page still hides the card, the speed measurement, and the model details here, and a policy that hid the Generation or Supervisor tab still hides the context length or automatic model management.

Tune settings automatically (off by default) lets AI:GO tune local models for Chat while you are away from the computer. It starts only after ten minutes with no keyboard or pointer input and no request through the router, with nothing loading or measuring, on AC power, and with memory to spare for the model. On Linux it needs GNOME or xprintidle to tell how long the computer has been idle, and without either it never starts. It tunes one model at a time, the running ones first, and only a model whose chat tune has something to try; a model is tuned again only after its file, the engine, or the computer changes. It runs the same measurement as Tune on the Benchmark page without the drafter comparison, stops the running models without asking, and runs them again afterward. Anything you do stops it at once and changes nothing: using the computer, a chat request, loading a model, starting your own measurement, switching to battery, turning the switch off, or quitting the app. A finished run applies only speed settings (batch sizes, threads, flash attention, GPU layers, CPU MoE layers, the MLX prefill chunk, and n-gram speculation), only when they are at least 1.1 times faster, and never a value you set yourself. It does not run the model again, so the values take effect on its next run. Applied values read Set by AI:GO in the Hardware tab of the model's settings and under Profiles in its detail, where Undo restores what was there before. The Benchmark page's History marks these runs Automatic and says why a stopped one stopped, and the line under the switch says what it is doing. A headless server never tunes on its own: the setting is stored there but does nothing.

Tune, which finds faster settings for one use when you ask, is on the Benchmark page (see Tune). Each model keeps one profile per use under Profiles in its detail here, where Undo restores the settings from before an apply (it asks first and names the date), and a saved profile's Apply opens its result on the Benchmark page. Apply and Undo are unavailable while another measurement runs, since both run the model again.

On a laptop, Defer while on battery holds the first-run check until the model's next run on power.

After an engine or app update, a model's number reads engine changed until it is measured again. Nothing measures on its own, since a check would interrupt a running model; instead the app keeps timing real use. Once ten chat answers of at least 64 tokens have run on the new engine with the same settings as the baseline, it compares their median answer speed with the baseline's, and below 80% of it the model's row and detail warn, for example "Qwen3-8B is slower than before the update. Answers 34 tok/s, previously 45 tok/s.", and one notification offers Measure again and Tune, once per model and engine build. Measure again runs the quick check and makes its result the new baseline either way: within 10% of the old speed it says the measurement matches before, so the slowdown may come from the conversations themselves, and otherwise the warning stays, and Tune in the notification opens the tune on the Benchmark page. Answers given with other settings than the baseline's are never compared, because a different setting is not a slowdown, and provider models are never compared at all. A headless server does the same with the chats its web UI runs.

Open Settings > Tools & extensions.

General holds the tool-calling switches and Available tools, which opens the list of tools with a switch and a Test action on each. The MCP servers and the web search tool follow it. Workspace context, Python sandbox, Local network access, and Skill shell injection each open from Advanced at the end.

Tool approval rules, under Allow auto-approval of tool calls, opens the rules that decide when a tool asks before it runs, with each rule's editor one level further in. This was a tab of its own, and an organization policy that hid it (policies) still hides it. See Tool Calling.

Allow auto-approval of tool calls and the Always allow this tool list are grants: they let a tool that asks by default run without asking. On the desktop you write them yourself; over the Management API they are written through PUT /api/v1/tools/grants, which needs the tool_approve scope, and PUT /api/v1/settings or a settings restore refuses to turn them on from a key without it. Turning them off needs only settings_write. A settings restore and PUT /api/v1/acp/settings apply the same rule to ACP permissions: a wider defaultPolicy, an autoAllow kind added, or an alwaysAsk kind removed needs tool_approve, while narrowing does not. See Grants.

This covers AI:GO as an MCP client, connecting to remote MCP servers. To expose AI:GO itself as an MCP server, see MCP Endpoint under Settings > API Server > Management API.

Voice input behavior

Open Settings > Voice Input.

Transcription model lists the Whisper models in your library with their sizes. When the library has none, a button downloads the model that suits this computer's memory: Whisper Large v3 Turbo with 12 GB or more, Whisper Small below that. The downloaded model becomes the transcription model when the download finishes. Choose a file sets a model file outside the library.

System tray and notifications

Open Settings > General. The System tray section shows or hides the tray icon (the menu bar icon on macOS) and, while the icon shows, sets what closing the window does: quit the app or keep it running in the tray. The tray settings were on the Appearance tab before, and a link or a search result for one of them now opens General. The Notifications section, in the desktop app, has one switch: Notify when a download or task finishes. While it is on, AI:GO sends a system notification when a model download or a Cowork task finishes and the app window is in the background. A task you stopped yourself is not announced, and neither is one that finishes while you are looking at the window.

Terms, privacy policy, and app information

Open Settings > General and scroll to the About section at the bottom. What's New shows the running version and opens the notes for it, and the section also opens the Terms of Service, the Privacy Policy, the open source licenses (Acknowledgments), and About AI:GO. Each one opens inside the Settings window, and the back arrow before its title returns to the list. About AI:GO holds the app's license, the manual, issue reporting, and the product tour. These links used to sit at the bottom of the sidebar. In the desktop app on macOS and Windows the Help menu opens the same documents, and on macOS About AI:GO is the first item of the AI:GO menu.

Squad run-history retention

Open Settings > Squad. Squad history retention sets how many completed executions each squad keeps (10 to 5,000) and can limit the storage each squad's execution logs, event ledgers, and reports take. A change applies when a run next completes, so lowering a limit does not remove anything at once. The section was on the Advanced tab before, and a link to it there now opens Squad.

Update channel

Open Settings > General, in the Updates section. The channel decides which build the update check fetches: Stable, Beta, or Canary.

Switching the channel moves you to that channel, and the app then offers the newest build of the target channel that was published after your current build, compared by publication date rather than version number. When nothing has been published after your build, the check reports no update and you stay where you are. Checks within the same channel keep comparing version numbers.

Connector configuration

Open Settings > Connectors. Each connector has a card with its status: Email, Google Calendar, Apple Calendar and Reminders, Microsoft 365, and Contacts, and a card opens the connector's settings. A card says Needs attention when a connection could not be restored after the app restarted, and its settings say what fixes it. Apple Calendar and Reminders, Microsoft 365 Connector, and Contacts walk through the last three. Once an account is connected, chat, agents, and automations can use it through its tools: reading runs without asking, and sending mail, changing or cancelling an event, and deleting ask for your approval every time.

Data Hub search presets and maintenance

Open Settings > Data, or the Data settings tab on the Data page. The tab opens on an Overview of what the Data Hub holds: the number of documents with a link to the Data page, whether Semantic search is ready (it shows Keyword search only until a preset names an embedding model), and the retry for unfinished imports. Search presets, watched folders, the scheduled URL re-fetch, wiki updates, and the trash follow. Citation export and the data doctor are under Advanced.

Each preset opens as its own page with its editor, its embedding run, and delete; going back with unsaved edits asks first. The Semantic search row at the top opens the preset whose model the embedding run uses.

Auto-extraction timings

Open Settings > Memory and, with auto-extraction on, select Tuning under Auto-extraction. It holds the extraction interval, the context window, the idle flush delay, the topic boundary gap, catching up on launch, and the topic-shift boundary signal. The Edit in Settings button on the Memory page's extraction status opens it.

Connection addresses

Open Settings > Advanced and, in the Network / Certificate section, select Connection addresses. It lists where each external service is reached: the update feed, the Hugging Face hub, the engine and runtime registries, web search, the messenger channels, and Microsoft sign-in and Graph. You can set a custom base URL there, such as a Hugging Face mirror; the Hub connection address row under Settings > Models > Hugging Face opens it too. An address with no certificate of its own uses the certificate in the same section. This was the Managed endpoints tab: an organization policy that hid that tab still hides it, and an address the policy manages is locked.

Supervisor history

Open Settings > Advanced and select Supervisor history. It shows the supervisor's status, its policies, recent decisions, and fallbacks, and, when the backend turns them on, predictive analytics, adaptive tuning, and webhooks. The supervisor's switch and behavior preset are Automatic model management in Settings > Models > Running on this computer. This was the Supervisor tab, and an organization policy that hid it still hides both. See Supervisor Agent.

Demo Mode

Open Settings > Advanced and select Demo Mode. See Demo Mode.

Agent registry sources

The sources that the community agents, squad templates, and community skills come from are managed on the Agents page: select Manage sources in its Community or Squad templates tab, or under Skills > Community. A link to the old Settings > Agent Registry tab opens them there, and an organization policy that hid that tab still hides them.

Agent Protocol (ACP)

Open Settings > API Server and, under Integrations, select Agent Protocol. It was a tab of its own; a link to that tab opens it here, and an organization policy that hid the tab still hides it. See ACP Server.

Maintenance tasks

Updates and privacy are in Settings > General. The network certificate, backup and restore, and reset are in Settings > Advanced.