Chat

Using ArchibotChat

Start threads, pick a context source, attach files, save prompts, and read responses on the Chat screen.

Customer adminsCustomer membersPlatform operators

Last updated June 18, 2026

The Chat screen lets you ask Archibot questions from the Console. The browser uses a hidden first-party credential, so you do not need an API key to chat. You only need an API key for scripts and integrations, which are covered in API keys.

If you are opening Chat for the first time, start with ArchibotChat getting started and ArchibotChat setup so access, credits, and your data-boundary acknowledgement are in place.

Start a thread

Each conversation is a thread. Threads keep their context source, attached files, and tool traces together.

1. Open Chat from the Console sidebar.
2. In the left rail, select an existing thread or choose New conversation.
3. Type your prompt in the composer.
4. Send the message with the round Send button or by pressing Enter.

If the left rail is collapsed, use the Threads tab on the left edge to expand it, or the history button in the header on narrow screens. The header shows the active thread title, the current context source, the message count, and how many files are attached.

When a thread has no messages yet, the empty state offers suggestion chips such as Explain an AXVW view, Find a stored procedure, and Review a workspace change. Selecting a chip drops a starter prompt into the composer so you can edit it before sending.

The composer

The composer sits at the bottom of the thread. Type your prompt in the text box, then use the icon row beneath it:

Enter sends the message and Shift+Enter inserts a newline. The Send button stays disabled until the composer has text, chat is enabled, and credits are available.

If chat is enabled but no credits remain, a banner appears above the composer with a Billing button, and the composer placeholder changes to Add credits to chat. See ArchibotChat billing and credits to top up.

Choose a context source

Context & tools controls what Archibot can read for the thread. Open it from the header button or by selecting the context label below the composer.

The dialog opens on the Context pill and lists the available sources:

• Static WebCentral — read-only files, the HQ database, and managed skills. This is the default and is always ready.
• Reference authoring — AVS working-file sessions in the reference WebCentral workspace.
• Workspaces — your running workspaces, each shown with owner, agent, and status. Pick one to let Archibot reason about that workspace’s runtime context.

Summary tiles at the top of the dialog show the selected context, the number of ready workspaces, attached files, and recorded trace events. Use the refresh control if a newly started workspace does not appear yet. To create or start a workspace, see Create the first workspace and Manage workspaces.

Attach files

Use attachments when the answer depends on a file Archibot cannot already read.

• Select the paperclip to upload from your device. The composer accepts PDF, spreadsheet (XLSX, XLS, CSV), drawing (DWG, DXF), Markdown, JSON, and image files.
• Select the archive box to open the Files tab, where you can attach an existing file, download, or remove it.

Attached files appear as chips above the composer. Remove a chip with its close button if the next prompt should not use it. Uploaded and generated files are also listed on the Files screen, where you can check expiry before relying on an older file.

Only attach data your organization has approved for hosted commercial SaaS processing through the configured model-provider path. Do not upload passwords, private keys, tokens, cookies, or unrelated personal information. Review approved data use before sending sensitive workloads.

Save and reuse prompts

The bookmark button stores prompts you reuse often:

1. Type a prompt in the composer.
2. Open the bookmark menu and choose Save prompt.
3. To reuse a saved prompt later, open the bookmark menu and select it; the body drops into the composer for editing.

Each saved prompt shows a short title and a preview of its body. Remove one with the trash control next to it. When you have not saved any yet, the menu reads No saved prompts yet. Saved prompts are stored in your browser, so they are local to the device and profile you are using.

Read responses

Archibot streams its answer into the transcript. A response may include:

• A direct answer with formatting and tables.
• Cited sources or referenced files.
• Generated files, shown as chips you can open or download.
• A Files for this turn row when files were attached or produced.
• Follow-up suggestions or warnings when information is missing.

While a response is streaming, the header shows a Streaming badge. To follow along, use the scroll-to-latest button if you scroll up during a long response.

Review any operational action Archibot drafts before you apply it. Archibot helps draft and analyze; your organization remains responsible for final decisions.

Inspect tools and trace

The Context & tools dialog has two more pills:

• Files — attach, download, or remove files for the thread without leaving Chat.
• Trace — the recorded tool steps for the latest response, grouped by tool with timing. If nothing has run yet, it reads No tool trace has been recorded for this conversation yet.

Use Trace to understand which tools Archibot called and whether any step failed. For an account-wide record of activity, see Activity and audit.

Header and chat menu

The thread header carries quick controls and a kebab menu (the three-dot button):

• Context & tools — open the context, files, and trace dialog.
• Export chat — download the current thread (download icon).
• Copy session ID — copy the thread’s session identifier (copy icon).
• New conversation — start a fresh thread (plus icon).
• Chat menu — the kebab opens shortcuts to Files, API keys, Activity, Billing, Support, and Setup, plus Export chat, Copy session ID, and Refresh session. It also shows whether chat is enabled and how many credits remain.

Credits and usage

Submitted prompts draw down credits. Streaming chunks, internal retries, tool steps, citations, browsing the Files list, and downloading an existing file do not draw separate charges. A prompt that fails before the server accepts it should not consume credits.

The kebab menu and the Context & tools summary both show remaining credits. For balances, top-ups, and budget caps, see ArchibotChat billing and credits.

Thread history and good practice

• Thread history is account-scoped; you see only threads your role and organization policy allow.
• Keep unrelated work in separate threads and let titles describe the task. This makes files and usage easier to review later.
• Keep prompts specific, and include the relevant date range, building, floor, room, asset, request, or person identifier.
• Attach only the files the question needs, and check file expiry before relying on an old one.
• Use API keys for scripts instead of trying to automate the browser chat surface, and revoke any key that may have been exposed.

If Chat will not send, the composer stays disabled, or context sources are missing, see ArchibotChat troubleshooting.