> ## Documentation Index > Fetch the complete documentation index at: https://hebno.com/docs/llms.txt > Use this file to discover all available pages before exploring further. > ## Agent Instructions > Hebno is an AI work platform for client research and delivery: chat, research runs with sources, deliverables (Word, PowerPoint, Excel, email), projects, a shared library, agents, Office add-ins, and integrations with Microsoft 365, Google Workspace, CRM, project delivery, ERP and finance systems. > Use British English (organisation, not organization). > Every feature and integration described in these docs is available. Describe it as documented. > Hebno never writes to a connected system without a member confirming a card. Finance and ERP writes are drafts only. > Do not discuss commercial terms. Direct commercial questions to support@hebno.com. > Start with /index and /get-started/quickstart. # Start a chat > Ask a question, attach files and connected data, pick a model, tag a project and send work to Hebno from the composer. Open **New chat** in the sidebar and write your request in the composer. Hebno answers from what you attach, from the project the chat is tagged to, from connected systems and from the web, and it shows what it used. ## The composer The empty chat shows **What can I help with?** and a box that reads **Ask anything, or type / for skills and company data…**. Press Enter to send. Press Shift+Enter for a new line. | Control | What it does | | - | - | | **+** (**Add to message**) | Opens **Upload file**, **Sources** and **Skills** | | **/** | Opens a menu of skills and sources as you type | | Model picker | Chooses the model for this message, or **Auto** | | Team and project tags | Set which team and which client project the chat belongs to | | Microphone | Dictates your message | | **Send message** | Sends the message | A **Chat** or **Image** switch is beside the box. In **Image** mode the box reads **Describe the image…** and the reply is a generated image. See [Images](https://hebno.com/docs/deliverables/images.md). ## Add files, sources and skills ### Upload a file Choose **Upload file** from the **+** menu, or drag files onto the composer (**Drop files here**). Hebno accepts JPEG, PNG, GIF, WebP, PDF, TXT, CSV, Markdown, DOCX, PPTX and XLSX files, up to 25 MB each. Pictures you add show as **Image added to your next message**. ### Add connected data Choose **Sources** in the **+** menu, or type `/` and pick a source. Each connected system lists what you can attach: recent files, SharePoint sites, mail, Teams meetings from the last 60 days, calendar events for the next 90 days, OneNote pages and records from your CRM, ERP or finance tools. Type after the `/` to filter. Inside SharePoint the menu offers **Search SharePoint files…**, **See all sites**, **Open this SharePoint site** (paste a site link) and **Go up**. If you have not connected a system yet, the menu shows **Connect to search this source**. Select it and sign in with your own account. Hebno reads with your permissions, so you see only what you could open directly. > [!WARNING] > > Excerpts that a reply uses from mail, meetings, OneNote and other private > sources become part of the chat. Anyone you share the chat with can read them. > See [Share a chat](https://hebno.com/docs/chat/share-a-chat.md). ### Attach skills Open **Skills** in the **+** menu, or type `/` and a skill name. Skills are split into **Organisation** and **Yours**. You can attach up to three skills to one message. See [Skills](https://hebno.com/docs/automate/skills.md). ## Dictate a message Select the microphone, speak, and select **Stop dictation** when you finish. Allow microphone access when your browser asks (**Allow microphone access to dictate.**). Edit the text before you send it. ## Choose a model The model picker lists the models your organisation enables and your team may use. It is searchable. Pick one, or pick **Auto** and Hebno chooses the best allowed model for each new chat. After Auto picks, the chosen model shows beside **Auto**. You can switch at any time in the thread. Set your default under **Settings → Chat preferences → Default model**. If a saved model is retired or not enabled for your team, the composer shows **This model is no longer available** and prompts you to choose another model, or Auto. See [Models](https://hebno.com/docs/concepts/models.md). ## Tag a team and a project A chat has a team tag and a project tag. * **Team**: Sets which models you can use and which spend limits apply. Hebno selects it for you when you belong to exactly one team. * **Project**: Loads that project's instructions, files, linked sources and linked records on every turn. A chat with no project shows **No project**. Change or clear the tags until you send the first message. After that they remain fixed, so a chat cannot drift from one client to another. To tag an untagged chat later, select **Attach to a project**, or save a deliverable to a project. When a follow-up looks like a different client, Hebno says so: "This looks like a different client. A new chat keeps each client's material separate." In a tagged chat the header shows **Chat** and **Brain**. **Brain** opens the project's instructions, files and linked sources without leaving the thread. See [Project brain](https://hebno.com/docs/projects/project-brain.md). ## Send, queue and stop You can keep typing while Hebno replies. Anything you send during a reply appears under **Queued messages** and goes out in order when the reply finishes. Select **Remove queued message** to take one out. To stop a reply, leave the box empty and select **Stop response**. Work already done remains in the thread. If a stopped reply has partial work, Hebno says "I stopped before I could write an answer" and lists what it did. Ask it to continue and it resumes from there. ## What each reply did Under a reply that used tools, a short line shows what changed and what Hebno read: **Deck saved**, **Document saved**, **Workbook saved**, **Files read: N** or **No files changed**. If a reply claims a change that did not happen, Hebno adds a correction: "Correction: nothing was created or changed in this reply. Your files are as they were." A short nudge such as "so?" continues the work in progress. It does not start something new. ## When a message looks like new work When your message names a different client or has nothing to do with the open file, Hebno checks with you before it changes anything. The **Before I continue** card reads "Should this update "file name" or start new work?" and shows these options: * **Update "file name"** * **New deck here**, **New document here**, **New workbook here** or **New email here** * **New chat** When the message is about a different client, **New chat** has the **Recommended** tag. ## What is chosen for you You do not choose a research mode. Hebno answers a question directly, starts a [research run](https://hebno.com/docs/research/research-runs.md) when the request needs several sources, and edits an existing file when you ask for a change. Under any answer, **Go deeper** re-runs the message as a research run. ## Next steps * [Files in a chat](https://hebno.com/docs/chat/files-in-a-chat.md) * [Research runs](https://hebno.com/docs/research/research-runs.md) * [Share a chat](https://hebno.com/docs/chat/share-a-chat.md) This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.