Quick Start

Use this guide when you want a working Dawn setup in one sitting. The goal is not to configure everything. It is to get one real agent, one real source, and one real conversation working end to end.

Before you start

• A Dawn account that can sign in to Dawn Web.
• One provider path ready for your first agent: OpenAI, Azure OpenAI, Anthropic, OpenRouter, or Codex after a ChatGPT link.
• One workspace owner who can configure shared channels, integrations, and access.
• One channel or surface you want to test first. Slack or Web Widget is usually the fastest first shared setup.
• Optional but recommended: one source such as a GitHub repo, Jira project, or Drive folder so the first conversation is grounded in real data.

Step 1: Create the workspace

1. Open https://app.dawnhq.ai and enter Dawn.
2. Create a workspace with a name that matches the team or function that will own it.
3. Confirm the workspace opens with the core tabs you need: Agents, Channels, Integrations, Memory, and Skills.
4. Open Workspace settings so you know where access and auto-join policy live before you invite anyone else.

Step 2: Create your first agent

1. Open Agents and create a new agent profile.
2. Choose the provider and model you actually want to run in production.
3. Write a short system prompt that states the role, boundaries, and expected style.
4. Set memory access intentionally: none, all workspace memory, or tag-scoped memory if the workspace will support multiple domains.
5. Leave this as your default first agent until the rest of the setup is working.

Use one agent per workflow, not one agent for everything. A support agent, docs agent, and release agent should usually be separate profiles with separate bindings.

Step 3: Connect one surface

1. Open Channels and add one shared binding such as Slack, Web Widget, or Telegram.
2. Choose a safe trigger mode. Mention-driven or DM-first behavior is the right default for most teams.
3. Choose a context mode. If you want Dawn to use surrounding discussion later, pick the mode that lets it use recent messages it actually receives.
4. Attach the binding to the agent from Step 2.
5. Send a real test message from that surface and confirm the thread appears in Dawn.

Private external chats are a separate surface from shared bindings. Slack DMs, Telegram private chats, Teams personal chats, and Discord DMs can stay personal while still using the same workspace context after identity linking.

Step 4: Link one real source

1. Open Integrations and connect one provider.
2. Validate credentials before linking any sources.
3. Link one narrow source set such as one repo, one Jira project, or one drive.
4. Assign that source to the same agent you created earlier.
5. Pick a source item with a known answer, such as a repo, ticket, or page title. Ask about that exact item and confirm the reply names or cites the linked source instead of answering generically.

Step 5: Add one skill and one memory page

1. Create one memory page for a stable team rule, checklist, or reference that the agent should keep using.
2. Create one skill for a reusable instruction such as reply structure, tone, or review format.
3. Assign the skill to the agent.
4. Run one test where the content should come from the linked source and the output shape should come from the skill, for example a checklist or structured summary.

If you need a one-off manual behavior, keep it as a manual skill and invoke it with #skill:slug in the message rather than turning it into an always-on default.

Step 6: Check access and account-level setup

1. Open Workspace settings → Access and invite one teammate or manager.
2. Decide whether Auto-join linked identities should be enabled for this workspace.
3. In Dawn Web chat, type /help and /whats_new so you know the current command surface is available.
4. If you want Dawn directly in VS Code, install the VS Code Extension. If you want Dawn available through Codex, Cursor, or GitHub Copilot, open Account settings → API credentials and create a personal MCP token.

If you want the command list and current behavior, use Slash Commands.

Production readiness checklist

• The agent replies in the intended channel, thread, or chat surface.
• The answer clearly references the linked source when the prompt depends on source data.
• Attachments and files work in the surface you plan to use.
• Memory and skill behavior are visible in at least one real conversation.
• Another user can access the workspace with the role you intended.
• If you use private external chats, the linked-identity flow is tested before rollout.

Next: use Channels, Integrations, Workspace Access, VS Code Extension, and MCP and API Credentials to harden the setup.

Validate this workspace setup and tell me what is missing for first production use.

Send me a short first-day operator checklist for channels, integrations, memory, and access.