Chat

Chat is the primary way you interact with agents in Sprigr Team. Whether you are asking a question, delegating a task, or reviewing a report, it all happens through the chat interface.

Starting a conversation

There are two ways to start chatting with an agent:

From the Chat page

Navigate to the Chat page from the sidebar. You will see a list of agents available to you. Click on any agent to open a new conversation, or select an existing conversation from the sidebar to continue where you left off.

From the agent detail page

Open any agent from the Agents page and click the Chat button. This opens a new conversation directly with that agent.

Try asking your agent

“What’s on the schedule for today? Any urgent jobs I need to know about?”

“Draft a quote for a hot water system replacement at 42 Smith Street. Standard Rheem 250L electric, including installation and old unit removal.”

“Check if we have any customers in the Southside area who are due for their annual gas safety check.”

The chat interface

The chat interface is split into two areas:

Conversation list (left sidebar)

The left side lists your conversations with the current agent, split into Pinned and Recent. Each entry shows a preview of the last message and the time of the most recent activity.

Click on any conversation to switch to it. You can have multiple conversations with the same agent — for example, one for an ongoing project and another for a quick question. From the list you can also:

• Search conversations by title.
• Pin a conversation so it stays at the top.
• Rename a conversation.
• Organise conversations into folders and filter the list by folder. Folders are yours; they do not change anyone else’s view.

Drafts and queued sends survive switching conversations, so you can move between chats without losing what you were typing.

Message area (main panel)

The main panel is where the conversation takes place. At the bottom you will find:

• Message input — A text box where you type your message. You can write anything from a short question to a detailed brief. Press Enter to send, or Shift+Enter to add a new line.
• Send button — Click to send your message.
• File attachments — Drag a file onto the composer or use the attach control to include images, PDFs, CSVs, and documents with your message.
• Mentions — Type @ followed by an agent’s name to bring another agent into the conversation. A dropdown lets you pick from the agents available to you.
• Stop — While the agent is responding, you can stop the current turn before it finishes.

Messages appear in chronological order with your messages on one side and the agent’s responses on the other. The agent’s reply streams in as it is generated, so you see it build word by word rather than waiting for the whole response. If you send another message while the agent is still replying, it is folded into the same reply rather than arriving late.

Effort and Plan Mode

Two controls in the composer toolbar shape how the agent works on a turn:

• Effort — A sticky per-conversation setting that controls how much reasoning and tool use the agent spends on each reply. Auto uses the agent’s default; the five explicit levels are Quick, Balanced, Thorough, Extended, and Maximum, from fastest and cheapest to deepest. The level persists for the conversation and is clamped to what the agent’s model supports.
• Plan Mode — A toggle that puts the agent in read-only mode: it gathers information and produces a plan but does not execute any action that would change data. A banner reads “Plan Mode — Agent will gather information and create a plan, not execute actions” while it is on. Turn it off to let the agent act.

Voice chat and hands-free mode

You can talk to your agent instead of typing. The microphone button dictates into the composer. Speak replies aloud has the agent read its replies back to you. Hands-free voice conversation mode goes further: it listens, sends automatically after a pause, and speaks the reply, so you can hold a back-and-forth without touching the keyboard. It ignores background noise and the agent’s own voice, and press Esc (or the exit control) to leave voice conversation mode.

Message types

You will see several different types of content in a conversation:

Your messages

These are the messages you send to the agent. They appear in the conversation exactly as you typed them.

Agent responses

The agent’s replies, which may include formatted text, bullet points, tables, and links. Agents format their responses to be easy to read and scan.

Tool use indicators

When an agent uses a tool during its response — such as searching a knowledge base, looking up an email, or calling an API — you will see an indicator showing what tool was used and a brief summary of the result. This gives you visibility into what the agent is doing behind the scenes.

For example, you might see:

• “Searched Shared Knowledge for warranty policy”
• “Sent email via Gmail to supplier”
• “Looked up job #4521 in simPRO”

These indicators help you understand how the agent arrived at its answer and confirm it used the right sources.

Knowledge search and content drafting

Agents can search your knowledge bases and use the results to create content. For example, asking your companion to “search our knowledge base for our refund policy and draft a customer email” will trigger a knowledge search, then compose a formatted email based on what it finds.

Delegation to other agents

Your companion agent can delegate tasks to specialist agents. When you ask it to check support tickets or handle a task outside its scope, it routes the request to the right agent and brings the results back to you.

The companion collects responses from the delegated agents and presents a consolidated summary with actionable items.

Artifacts

When an agent produces content such as a document, a table, a chart, a rendered HTML page, an image, or code, it appears as an artifact you can view inline. A collapsible artifacts panel at the top of the conversation collects everything generated in that conversation, grouped into All, Images, Code, and Docs so you can find earlier outputs quickly.

Reasoning

For turns that involve deeper thinking, the agent’s reasoning is shown in a collapsible disclosure above its reply. Expand it to see how the agent worked through the request before answering.

Decisions

When an agent needs you to choose between a few options, it shows tappable buttons instead of asking in prose, and waits for your answer before continuing. See Decisions and approvals for how these work across channels.

Marketplace app cards

Installed marketplace apps can render interactive cards directly in the conversation, for example an order, a ticket, or a settings form from the app. Cards update in place as things change instead of stacking duplicates, and they have a mobile-friendly layout. Where an app action needs a person’s sign-off, the card is where you approve it; see Decisions and approvals.

Pull request cards

When an agent works on a GitHub or Sprigr-hosted repository, a live pull request status card appears in the chat showing the PR state (Open, Draft, Merged, or Closed), its CI checks, and a View toggle that opens a diff viewer with per-file patches. From the card the agent can review, watch checks, and merge on request.

Browser sign-in and co-browsing

If an agent needs to sign in to a password-protected website on your behalf, it shows a secure connect chip in the chat. You type the credentials into the chip; they go straight to encrypted storage and the agent never sees the password. While the agent works a website you can watch its browser in real time in the Live browser view panel, Take control to click or type yourself, and Return control to hand it back.

Building dashboards and websites

Agents can generate entire web pages and dashboards from a natural language description. Ask your companion to create a dashboard with specific metrics, charts, and styling and it will write the code and render it inline in the conversation, where you can preview it straight away.

To publish something for others to reach, agents can also deploy sites to Sprigr’s hosting. Deployed sites are managed on the Websites page, where you can view them, see their activity, and roll back to an earlier version.

Iterative refinement

You can refine anything the agent creates by giving feedback in natural language. The agent reads the source, understands what needs changing, and produces the update, all within the same conversation.

Conversation history

Every conversation is saved automatically. You can return to any previous conversation from the sidebar and pick up where you left off. The agent remembers the context of the conversation, so you do not need to repeat yourself.

Conversations persist across sessions. If you close your browser and come back later, all your conversations are still there.

Note

Each conversation maintains its own context. If you told the agent something in one conversation, it will not automatically know about it in a different conversation. Start a new conversation for a new topic, or continue an existing one if you want the agent to remember what was discussed.

Tips for better conversations

Get the most out of your agents

• Be specific — “Send a summary of overdue invoices from March to the finance team” works better than “Help with invoices.”
• Provide context — If you are referencing a specific project, customer, or document, mention it by name so the agent can look it up.
• One task at a time — Agents handle focused requests better than a long list of unrelated tasks in a single message. Break complex requests into steps.
• Correct the agent — If an agent gets something wrong, tell it. It will adjust and improve within the conversation.
• Ask follow-up questions — Agents maintain context within a conversation, so you can naturally build on previous messages. “Now break that down by month” works perfectly after a summary.

Next steps

• Inbox — View all your conversations in a thread-based inbox format.
• Team Rooms — Talk to your team and your agents together in DMs, groups, and channels.
• Channels — Connect agents to Telegram, WhatsApp, or Slack so others can chat with them too.
• Decisions and approvals — Answer an agent’s questions with tappable buttons on any channel.
• Companion Agents — Use your personal companion agent as a single point of contact that delegates to specialists.
• Knowledge Bases — Make sure your agents have access to the information they need to answer accurately.