Atlasnote ProConfigure your first AI model
Guide navigation

Type to search. Use ↑ ↓ to select, Enter to open, and Esc to close.

Meet Atlasnote ProGetting started · Understand how local sources, the AI workspace, projects, Wiki, and outputs fit together.
Install, register, and updateGetting started · Install on a supported Mac, create your account, and understand application updates.
Configure your first AI modelGetting started · Save a provider, choose a chat model, and check the connection with a short request.
Complete your first taskGetting started · Produce a source-based summary, check a citation, and reopen a saved output.
Find your way aroundGetting started · Locate Home, Library, Folders, Projects, Code, Wiki, Skills, and Plugins.
Start and continue conversationsAI workspace · Describe a goal, choose how to handle it, and continue work in an existing conversation.
Select sources and search your libraryAI workspace · Control the material available to a task and distinguish attachments from saved sources.
Set task permissions and stopping conditionsAI workspace · Choose file and network access and set optional limits for a task.
Track, cancel, and recover tasksAI workspace · Check execution, goal acceptance, and file delivery separately.
Import webpages, files, and notesCollect and read · Add material to your Library and check that its content is readable.
Read, annotate, and translateCollect and read · Keep notes close to the source and check translations against the original.
Read audio, video, and transcriptsCollect and read · Import media, inspect transcripts, and navigate between text and playback.
Organize sources and foldersCollect and read · Use filters, folders, and project associations without confusing organization with file moves.
Create a project and organize workProjects and research · Keep related sources, documents, and tasks together.
Run deep researchProjects and research · Define scope, select local or online sources, and inspect the report's evidence.
Verify citations and sourcesProjects and research · Return to original passages and distinguish evidence, inference, and unknowns.
Build a knowledge Wiki from sourcesKnowledge Wiki · Organize selected material into readable, traceable, maintainable knowledge pages.
Review Wiki changes and maintain evidenceKnowledge Wiki · Inspect differences and sources before applying knowledge updates.
Write, edit, and refine documentsCreate and deliver · Draft from sources, define the scope of changes, and retain traceable outputs.
Generate summaries and mind mapsCreate and deliver · Extract key points and structure, then check coverage against the original.
Create podcasts, presentations, and mediaCreate and deliver · Review the script or outline, configure required services, and inspect actual media files.
Save, export, and inspect output versionsCreate and deliver · Confirm file delivery, continue editing, and retain versions and sources.
Discover, import, and use skillsSkills and automation · Bring reusable methods into a task while reviewing their source and capabilities.
Turn completed work into a reusable skillSkills and automation · Draft a method from completed work and validate it on another input before scheduling.
Create and manage scheduled tasksSkills and automation · Set frequency, time zone, source scope, and authorization, then inspect actual runs.
Use Dream overnight organizationSkills and automation · Choose a model and local schedule, then review the resulting suggestions.
Connect plugins and MCP toolsAdvanced use · Add a connection, test available tools, and select it for a task.
Browse code and use CodeModeAdvanced use · Connect read-only source code and understand optional script-assisted processing.
Connect an external AgentAdvanced use · Check local installation and sign-in, then select an external executor.
Send tasks through messaging channelsAdvanced use · Connect a channel, approve a controller, and inspect execution on the desktop.
Configure models and external servicesSettings and data · Manage providers, text channels, local endpoints, and media generation separately.
Understand local data and AI authorizationSettings and data · Distinguish local storage, model processing, source sensitivity, and file access.
Back up, recover, and move your librarySettings and data · Create a checked backup and distinguish moving, reconnecting, and cloning.
Migrate and import conversation historySettings and data · Export inspectable history and preview external conversations before merging.
Inspect usage and control costsSettings and data · Distinguish reported model usage, unknown records, and task cost estimates.
Troubleshoot common problemsHelp and troubleshooting · Check configuration, sources, execution, and saving according to where the failure occurs.
Supported formats and scopeHelp and troubleshooting · Check device requirements, file limits, processing needs, and version boundaries.
Contact support and report a problemHelp and troubleshooting · Provide enough information to reproduce an issue without exposing private data.
Tutorial: save a summary with sourcesTutorials · Ask about a short source, verify the answer, and deliver a saved file.
Tutorial: produce a multi-source research reportTutorials · Compare sources, preserve disagreements, and deliver inspectable conclusions.
Tutorial: build and update a Wiki pageTutorials · Create a knowledge page from sources, then review an evidence-based update.
Tutorial: turn sources into an article, podcast, or presentationTutorials · Reuse a checked outline and inspect the delivered format.
Release notes and documentation updatesRelease notes · Check the current installation version, upgrade instructions, and documentation scope.

Configure your first AI model

Save a provider, choose a chat model, and check the connection with a short request.

Applies to · 0.0.2Content reviewed ·
On this page

For initial setup, get one chat model working. Configure speech, image, and video services when you need them.

Screenshots show the Chinese application interface in a Test environment with example data.

Prepare your service details

Have your provider's API key and a model name available to that account. The provider charges for usage. A chat subscription and API access may be separate products; a subscription alone does not establish that your API key will work.

Add and select a model

During initial setup: select a cloud service in the AI connection step, enter the provider, model, and API key, then save and test. Resolve any test feedback before continuing.

First AI connection: provider, model, and key must agree. This example has no key entered and has not passed a connection test.

  1. 1Choose your provider and a model available to your account.
  2. 2Enter the API key for that service.
  3. 3Save and test, then check the result.

Test environment · Example data · Chinese application interface

Initial setup: select a cloud model and test the connection

First AI connection: provider, model, and key must agree. This example has no key entered and has not passed a connection test.

For a local model: switch to local setup and start Ollama, LM Studio, or your compatible service. Check the address, read the available models, and select one. Verify tool-call support before relying on it for tool operations.

For a local model, start the local service before reading its available models.

  1. 1Switch to local-model setup.
  2. 2Check the service address and read its models.
  3. 3Confirm tool-call support only when the model provides it.

Test environment · Example data · Chinese application interface

Initial setup: read available models from a local service

For a local model, start the local service before reading its available models.

After reaching the main interface: you can add or change a connection in Settings at any time:

  1. Open Settings with the sidebar's lower settings icon or profile menu, then choose API.
  2. Add a provider and enter your key. For a custom service, also check its API address and interface type.
  3. Save the provider and check the confirmation.
  4. Open model channels and select that provider and an available model for the chat channel.
  5. Run the channel probe, then start a new conversation and send a short question.

Channel changes save automatically and affect new conversations. Existing conversations retain their model selection. Saved keys are not displayed in plaintext; an empty edit field does not mean the stored key is missing.

API settings before any provider is added. Add a service before assigning models to features.

  1. 1Add a preset provider.
  2. 2For a custom service, check the address and interface type.

Test environment · Example data · Chinese application interface

Add a model-service provider in Settings

API settings before any provider is added. Add a service before assigning models to features.

Confirm success

The model probe should not report failure, and a new conversation should return a normal response. Test a source-based question next, so model connectivity, complex file processing, and media generation do not all need troubleshooting at once.

Model channels are configured per purpose. The test-model name is a fixture and its status is unprobed; use a real model from your service.

  1. 1Choose a provider and model for Chat.
  2. 2Run the probe and check the actual connection result.
  3. 3Check privacy policy, content level, and tool-call capability.

Test environment · Example data · Chinese application interface

Model channels: choose a chat model and run its probe

Model channels are configured per purpose. The test-model name is a fixture and its status is unprobed; use a real model from your service.

You can skip AI configuration to read local sources and configure a model when you need generated answers.

If the connection fails

Check the key, service address, model name, and provider account allowance. When switching providers, choose a model supported by the new provider. See model and service configuration and troubleshooting.

Next step

Complete your first task by importing a source and asking a question with citations.

Configure model, media, and connected services as needed. See each article for requirements.

Get help →