Working with agents

You interact with Devs.ai through chat: ask questions, work with files, and run tools. Start from the homepage with the default chat experience, or open an agent you created or one that is shared with you.

If you already have ongoing conversations, open Chat to view and manage them. Each chat is a single conversation thread with its own context. You can continue a conversation as long as needed and return to previous chats at any time.

Chat runs on a conversation engine tuned for longer threads, tools, and streaming. Responses tend to stream more smoothly, and if you refresh the page or lose connection briefly mid-reply, you are less likely to lose the whole answer. If something still looks wrong, try again after a short wait. Upstream model providers sometimes have incidents.

📝 Note: The new chat workspace (home shortcuts, canvas, live browser panel, and related tools) rolls out as it is enabled for organizations. If you do not see these options yet, you still have classic chat. Available features also depend on your organization’s settings, permissions, and plan.

Chat workspace home

When the new chat workspace is enabled, the home screen asks what you want to work on and offers shortcuts so you do not have to know which tool to pick first. Typical shortcuts include:

Shortcut	What it starts
Write a document	Drafting in an editable canvas beside the chat
Browse a website	A live browser task you can follow in the side panel
Create an agent	Building a reusable agent from a description in chat
Create an image	Image generation in the side panel, with room to refine
Analyze files	Working with uploads you attach
Research a topic	Deep Research for a broader evidence-based report

Suggested prompts and examples appear under the shortcuts. You can also type freely in the message box (Describe what you want to do…).

During a conversation, a side panel keeps the active document, image, or browser task in view. You can drag the divider to resize chat and panel. A strip of items created or used in the chat lets you reopen documents, images, browser sessions, and sources. Tool activity shows recognizable connector names and short descriptions of what is happening.

The same workspace pattern extends to chats inside projects.

Where the updated activity view is enabled, one expandable row follows thinking, tool use, and writing, with clear states for stopped work, errors, and requests for your input. Sources, approvals, and browser actions remain accessible from that activity.

Queue your next instruction

If you have the next instruction ready before the answer finishes, use Add to queue. Queued prompts send automatically when the assistant finishes, in order.

• Edit or Remove a queued message, or Clear all.
• Send now sends immediately and interrupts the current response (there's also a keyboard shortcut for this).
• Stopping a response can pause the queue. Use Resume when you are ready to continue.

Canvas: write and refine documents

Ask chat to draft a document (or use Write a document). An editable canvas opens beside the conversation so you can shape the result while you keep chatting.

• Edit directly in the rich-text editor—headings, lists, tables, text colors, and highlights.
• Select a passage and use Ask for a change (for example make it shorter, more formal, fix grammar, add detail, or describe a custom change). The agent works from the current document, including your own edits.
• Add images, diagrams, and charts when the draft calls for them. You can resize images in the editor.
• Version history keeps earlier drafts so you can revisit a previous version or jump back to the message that produced it. History identifies who created each version when the document is shared.
• Export as Markdown (.md), text (.txt), Word document (.docx), or Open in browser (for print / save as PDF) for a formatted PDF. Download original downloads the Markdown source.
• Share — Document owners can create an organization-wide view link for signed-in colleagues, and disable that link later. Where canvas sharing is enabled, owners can also invite named collaborators (see below).

Export to Word

Choose Export → Word document (.docx) to open the canvas in Word or Pages. Text, headings, lists, links, colors, highlights, and tables stay editable. Images, charts, and diagrams travel as embedded images. Export includes what is currently in the canvas, including edits you have just typed.

Word export is available from chat canvases, collaborative documents, and organization share links, according to your access. If you ask chat for a Word copy of an existing canvas, it can point you to this control—you start the download from the document you already have.

Share a canvas with your team

Bring colleagues into a document while keeping each person’s drafting conversation private. Named collaboration is available where canvas sharing has been enabled for your organization; organization-wide view links continue to work separately.

Canvas owners can invite named Viewers and Editors:

• Viewers can open and read the shared document.
• Editors work in a private draft, ask the AI for changes in their own chat, review the differences, then choose Apply to shared canvas. Only the document changes they apply become shared. Prompts and AI replies stay in that person’s chat.

Invitations can send an email with a direct link. Sharing notifications can be managed in account settings. Collaborators find shared documents under Library → Shared with me, with their role on the card; owners can see which documents they have shared. See Files library.

Shared updates appear in open canvases without a full page reload. Manual edits save after a short pause, and refreshes preserve work you are still editing. Conflicting changes cannot silently overwrite a newer shared revision.

Browse websites alongside the AI

Ask Devs.ai to visit a site, explore pages, and compare what it finds. The browser panel shows the agent's work as it happens. When the task finishes, a compact summary and screenshot strip help you revisit the pages it explored. Browser sessions also appear in the item strip so you can reopen them.

If a site needs sign-in, the browser can pause on Sign in securely—a dedicated form that sends credentials straight to the browser, not into the chat transcript. Values are cleared after a short time (for example 15 minutes). You can continue or Skip. Individual websites can still restrict automated browsing.

Browsing in default chat is related to the agent Browser tool; for enabling Browser on a custom agent, see Configuring extended capabilities.

Deep Research

For work that needs a broader search and a closer review of the evidence, start Deep Research from the chat controls. You can also use the home Research a topic shortcut, or an explicit request in your message (for example “Use deep research to…”).

An in-chat progress card follows planning, searching, reviewing, and writing. Expand the details to see searches, then use Open report to read the result in canvas. Reports include sources and call out gaps or uncertainty when the evidence is incomplete. Research keeps the requested report language when it draws on sources in different languages.

You can leave the chat and return later. Completion emails can notify you when the report is ready, according to your notification settings. Longer runs have dedicated time to write after searching, with clearer outcomes when a run is paused, stopped, or unable to finish.

From a finished report you can keep editing in canvas, ask for a shorter version, turn findings into a plan, or continue into related flows such as a presentation when App Builder offers that path.

Create an image

Use Create an image or ask in chat. The image appears in the side panel. You can refine it—including selecting a region and asking for a targeted change—adjust aspect or style when those controls are available, and download the result (for example as JPEG). Generated images also show up among the items you can reopen from the conversation.

Create an agent from chat

You can create and configure an agent directly in default chat instead of opening the agent builder first. Describe the job, how it should respond, and which resources to use. Devs.ai can save the agent and offer Open agent and Try agent.

Keep refining in the same conversation—instructions, model, tools, knowledge, skills, and inputs. Follow-up changes update the same agent. Existing permissions and connector setup still apply. For create-from-chat steps, Simple mode, and the Advanced editor, see Creating agents.

Stay oriented in long replies

Scrolling a long answer should not leave you guessing which question produced it. Your prompt stays pinned to its turn while you scroll through the response, in a compact single-line form that does not crowd the answer. Message timestamps show a short relative time in the thread. Hover a timestamp to see the full date and time.

In chat options, Thinking and Reasoning Effort settings appear only when the selected model supports them. If those controls are missing, the model simply does not expose that capability.

For optimal results, try to frame prompts that are aligned with the knowledge bases or data stores you have configured. In case you don't get the intended response, you can try new inputs and instructions. Through such engagement, agents constantly learn and add new information to their knowledge base.

Using agent capabilities during a conversation

An agent’s responses depend on how it has been configured. Based on enabled tools and data sources, an agent may ask follow-up questions, reference connected data sources, or perform actions using tools.

You can upload files directly into a chat. Agents can use these files to answer questions, summarize content, or perform analysis, depending on the tools and data access configured for the agent. Uploads stay private to your organization; if you open or download a file later, access is time-limited rather than a permanent public link. Choose From Library to attach files you already generated or uploaded—see Files library. When web search is enabled for the chat or agent, answers that use the web include citations to the real source pages, with duplicate links collapsed.

From an active chat you can also access additional options such as sharing the agent with others, resetting the chat context, and deleting the chat.

Continue in App Builder

When a request is really a build (an app, site, tool, or presentation), chat may show Build this in App Builder? (or Build this deck in App Builder?). Choose Open App Builder to continue with the starting prompt and optional Files to bring along, or Stay in chat to keep working in the conversation. Details are in App Builder.

Connect tools while you chat

Chat can help you discover a relevant connector even if your organization has not enabled it yet. A setup card can offer Enable, Install, or Finish setup for admins, or guide members to ask an admin. When a matching connection is already available, chat prefers it for supported tasks instead of browsing the service in the live browser. A connector your organization has already configured can also appear with Enable in chat so you can select it for the next message (complete sign-in if needed)—separate from the org-level Enable in Chat toggle on Connectors. Saved-agent chats continue to use the tools configured for that agent. In account connector settings, use Disconnect or Reconnect when a connection needs to be refreshed. Organization admins manage the catalog in Connectors.

Schedule work from chat

You can ask Devs.ai to create or manage a routine from a regular chat or an agent chat—describe the task and timing, then refine it in the same conversation. When a turn looks like something you might repeat, chat can offer Make this a routine? Choosing Set up routine starts the setup discussion; the offer itself schedules nothing. Calendar bookings and spending limits appear where those capabilities are enabled for your organization.

Context usage in chat

While you chat, a context usage ring may appear in the message input area. It shows how much of the active model’s context window the current conversation is using.

Open the breakdown

Click the ring to open the Context usage popover. You see:

• How many tokens are in use versus the model’s real context limit for this thread, with the soft max (your budget percentage of that window) called out when budgets apply.
• A colored breakdown by message type (for example user messages, assistant replies, tool results, and attachments).
• An Estimated label when the count is approximate rather than exact.

The ring and breakdown scale to each model’s actual window, so switching models updates what “full” means. The ring changes color as usage grows so you can spot conversations that are getting full.

Adjust your chat budget

In regular chat (not agent-backed threads with a fixed cap), the popover can include a control to configure your personal chat context budget—the percentage of the model window reserved for history on this surface. Use it when you want more room for the current thread or prefer to leave headroom for long tool outputs.

Organization administrators can set default Chat context budget and App Builder context budget percentages in organization settings. Your personal override applies on top of those defaults for chat.

Models with very small context windows use the full window automatically; the percentage control is hidden in those cases.

How this relates to compression

When a conversation approaches its limit, the platform may also summarize older messages automatically. Context usage shows live fill level; compression reduces what is sent to the model. Both work together so long threads stay usable.

Related topics

• Creating agents — create from chat, Simple mode, and the Advanced editor.
• Chat history compression — automatic summarization when conversations grow long.
• Files library — Library gallery and Account/Organization Files.
• App Builder — continue a build idea from chat when offered.
• MCP servers — connectors you can enable from chat or the Connectors hub.
• Skills — reusable instructions you can import or link while you work.
• Routines — schedule from chat, Calendar bookings, and spending limits.
• Projects — shared context across a group of chats.