Dawn Documentation

These guides are written for the people actually operating Dawn day to day: workspace owners, managers, and teammates. They explain what the product does today, how the current UI works, and what to validate before you depend on it.

Recommended setup path

1. Create your workspace and first agent

   Start with one clear agent profile, one workspace, and one owner who can manage channels, integrations, and access.

2. Connect one channel or chat surface

   Use a shared binding for team-facing conversations, or a private surface such as a DM if the conversation should stay personal.

3. Link one real source

   Connect one repository, tracker, or drive source and bind it to the agent before broadening access.

4. Add memory, skills, and access policy

   Decide what persistent knowledge the agent should use, what reusable instructions should shape replies, and who should be invited into the workspace.

5. Optional: use Dawn from desktop, editor, or terminal

   Install Dawn Desktop for a native chat client, use the Dawn VS Code extension for editor chat, use Dawn CLI for terminal work, or create a personal MCP API credential if you want Dawn available through Codex, Cursor, or GitHub Copilot.

What changed recently

• Workflows: Create event-driven agent runs with recipes, a visual canvas, validation, run history, and discussion handoff.
• Live Dashboards: Turn Dawn output into interactive dashboard views that can be revised, refreshed, and shared.
• Agent Gallery: Start from curated agent templates, then customize prompts, models, skills, and sources.
• Agent Context Optimization: Keep runs efficient with prompt caching, deferred tools, and native tool search where supported.
• xAI Grok Models: Configure xAI as an AI provider and select Grok models for workspace agents.
• Personal Connections: Connect personal Gmail, Google Calendar, Google Drive, and OneDrive accounts without making them workspace-wide sources.
• /reflect: Audit recent workspace context and suggest memory, skill, command, or workflow improvements.
• Memory Pages: Use workspace memory for shared team knowledge and personal memory for user-owned context.
• JavaScript Scripts: Let agents run controlled JavaScript for data transformation, file inspection, and generated outputs.
• Files and Attachments: Work with file mentions, generated files, previews, thread files, and durable attachments.
• Voice and Transcription: Use voice input, audio/video transcription, transcript files, and provider usage tracking.
• MCP Gateway: Connect hosted MCP servers as workspace integrations, link selected capabilities, and assign them to specific agents.
• FourA: Connect FourA so assigned agents can inspect public web pages and HTTP APIs during research and operations work.
• Dawn MCP: Create named API credentials, connect Dawn to editor clients, and use your existing workspaces and agents over MCP.
• VS Code extension: Use Dawn directly inside VS Code with private threads, workspace selection, agent selection, and current-workspace context.
• Dawn Desktop: Use Dawn as a native chat client with private threads, Desktop-enabled agents, local project context, attachments, and HTML artifact previews.
• Dawn CLI: Run private Dawn threads from the terminal with local workspace context, attachments, and non-interactive prompts.
• Browser extension: Use Dawn with the page you are viewing and carry web context into a private thread.
• Slash commands: Use product commands such as /help, /whats_new, and /link directly in supported chat surfaces.
• Command Reference: Open a page for each built-in slash command and shipped Prompt Command.
• Prompt commands: Turn repeated workflows such as PR review, ticket planning, release notes, and prototypes into reusable slash-triggered prompt presets.
• Identity linking: Link a private Slack, Teams, Telegram, or Discord identity to your Dawn account so Dawn recognizes you across channels and Dawn Web.
• AI providers and Bring Your Own Keys: Use row-based Bring Your Own Key configuration, per-provider model allowlists, custom model IDs, and a separate ChatGPT section for Codex.
• Confluence: Connect Confluence Cloud with your Atlassian site URL, email, and token, then link the spaces your agents should read.
• Notion: Connect a Notion internal integration, share root pages with it, and expose those page trees to agents.
• Figma: Connect Figma with a token or OAuth, then paste file or node URLs to link the design files, components, and variables agents should use.
• Framer: Connect a Framer project editor URL and Server API key so Dawn can discover the site, pages, and reusable styles.
• MailJet: Connect MailJet for contact list, template, send, and campaign operations.
• SendGrid: Connect SendGrid for contacts, dynamic templates, transactional sends, and Single Sends.
• Skills: Use global and workspace skills, invocation modes, and per-agent assignment to make outputs more consistent.
• Skill Reference: Open a page for each default skill Dawn ships.
• Agent-authored skills and commands: Let agents help create or update workspace commands and skills with reviewable audit history.
• Telegram: Current Telegram setup guidance covers webhooks, chat IDs, group context, and private conversation behavior.
• Jira: The supported Jira path is your site URL plus Atlassian email and a normal Jira Cloud API token.

Guides

• Quick Start: Run your first working Dawn setup in one session.
• VS Code Extension: Use Dawn as a first-party chat client inside VS Code instead of configuring an MCP client.
• Dawn Desktop: Use Dawn as a native desktop chat client with workspace switching, personal threads, local tools, and artifact previews.
• Dawn CLI: Use Dawn as a terminal client with private threads, local project context, attachments, and scripted sends.
• Browser Extension: Use Dawn against the current browser page from a first-party extension.
• Channels: Configure shared bindings, private channel behavior, and context mode correctly.
• Integrations: Connect code, ticketing, document, design, web data, email, and custom tool sources the way Dawn currently expects them.
• MCP Gateway: Connect hosted MCP servers when your team needs custom tools or long-tail systems inside Dawn.
• AI Providers and Bring Your Own Keys: Control owner-managed Bring Your Own Key rows, per-provider model availability, and billing behavior for workspace-backed runs.
• Workspace Access: Invite teammates and managers, review roles, and control linked-identity auto-join.
• Memory Pages: Manage workspace memory and personal memory separately.
• Skills: Understand scopes, priorities, overrides, and invocation modes.
• Skill Reference: Browse every shipped default skill page.
• Prompt Commands: Create reusable command workflows with global defaults, workspace overrides, markdown bodies, and revision history.
• Command Reference: Browse every shipped slash command and Prompt Command page.
• Workflows: Create event-driven automations with recipes, source events, agent-run steps, run history, and discussion handoff.
• Live Dashboards: Create interactive views from conversations, files, and connected-source data.
• Agent Gallery: Create agents from curated templates and validate them before rollout.
• Agent Context Optimization: Understand prompt caching, deferred tools, and native tool search behavior.
• JavaScript Scripts: Use controlled scripts for deterministic file and data work.
• Files and Attachments: Understand uploads, mentions, generated files, previews, and thread files.
• Voice and Transcription: Turn speech, audio, and video into reusable Dawn text artifacts.
• Personal Connections: Connect user-owned inbox, calendar, and drive context.
• MCP and API Credentials: Use Dawn from Codex, Cursor, or Copilot with your own account access.

Before you go live

• Confirm at least one active agent profile is bound to the channel or surface you expect users to use.
• Confirm the binding trigger mode and context mode match how your team actually talks to Dawn.
• Verify linked sources are assigned only to the agents that should use them.
• Test one reply with attachments or files if your workflow depends on them.
• Invite at least one non-owner user so you can confirm access behaves correctly for managers or teammates.
• If you use private external chats, test the linked-identity flow and decide whether workspace auto-join should stay off or be enabled.
• If you enable workspace AI providers, validate the effective credential mode before relying on Codex or Bring Your Own Key (BYOK)-backed runs.

Give me a readiness checklist for this workspace before going live.

List missing setup items across channels, integrations, and agent bindings.