# Web Chat

Chat with any AI employee directly in the workspace, with a streaming reply, a live execution timeline showing what it is reading, thinking, and doing, and inline approve/reject buttons when it needs a sensitive action confirmed. It is the default channel: nothing to configure, works on desktop or mobile, and every conversation persists so you can pick it back up later.

Chat is the primary way you work with your AI employee. Messages stream in real time so you see the response forming word by word. More importantly, you see the reasoning: what the employee is thinking, which tools it is calling, what decisions it is making, and why. This is not a chatbot that spits out a block of text. It is a colleague showing their work.

Imagine asking your employee to research competitors in your space. You watch it decide to search the web, pick specific queries, read through results, discard irrelevant pages, extract pricing data, and start composing a summary. If it heads in the wrong direction, you stop it mid-sentence and redirect. "Focus on European competitors only." It adjusts immediately and continues.

The chat supports rich media: images, files, tables, code blocks, and structured documents appear inline. Your employee can share a generated image, attach a spreadsheet, or display a formatted report directly in the conversation. Everything stays in one thread, and the full history is searchable.

Attachments travel the other way too. Drop an image, a PDF, a spreadsheet, or a document into the message and your employee reads it as part of the request. The first line of the reply names the file it opened, so you are never left watching a blank indicator wondering whether the upload actually arrived.

Conversations persist. Close the tab, come back tomorrow, open the same thread, and everything is where you left it, including the files you sent and the work that came out of them. If a message ever fails to leave your browser, it stays on screen with a retry instead of quietly vanishing, so nothing you typed is lost to a bad connection.

You decide where your employee should stop and check with you. Any tool can be set to ask first, and when the employee reaches that point it pauses inside the thread with approve and reject buttons rather than acting and telling you afterwards. Approve and it continues from exactly where it paused, with everything it had already worked out still intact.

## Real-Time Streaming with Visible Reasoning

Every conversation with your AI employee streams in real time, word by word, so you always know what is happening. Unlike black-box AI tools that return a finished answer with no explanation, Sistava shows the agent's full reasoning process as it works through your request.

Visible reasoning means you can catch mistakes before they compound. If the agent starts heading in the wrong direction, you see it immediately and can redirect mid-sentence without waiting for a wrong answer to finish.

## Tool Usage You Can Watch and Control

When your AI employee uses a tool, like running a search, calling an API, or reading a file, you see exactly which tool was called, what arguments were passed, and what came back. There is no guessing about what the agent actually did.

This transparency turns the agent into a collaborator rather than a black box. You can stop the agent at any point, override a tool call, or redirect the task entirely without losing context or starting over.

## Built for Iterative, Back-and-Forth Work

Sistava chat is not a one-shot prompt interface. It supports deep, multi-turn conversations where you refine, push back, add context, and guide the agent across dozens of exchanges.

Whether you are drafting a proposal, researching a market, or debugging a workflow, the AI agent maintains full context throughout the session. Each message builds on the last, and the agent adapts in real time as requirements evolve.

## How It Works

**A real-time chat interface with streaming output, visible reasoning steps, and live tool execution.**

Chat is the primary interface for working with your AI agent. Responses stream in real time so you see output as it is generated, not after a long wait. When the agent reasons through a complex problem, its thinking is visible inline. You can watch it plan, reconsider, and decide before it commits to an action. Nothing happens in a black box.

Tool execution happens live inside the chat thread. When the agent runs a web search, calls an external API, generates an image, or creates a document, the result appears in the conversation with full context about what was done. You can stop the agent mid-task at any point and it halts cleanly without corrupting its work. Every conversation is logged, searchable, and reviewable.

Chat, mobile, voice, and every other channel run on the same underlying stream, which is why a conversation you started on your phone continues in the browser with nothing missing. The employee's status, the files it produced, and any approval it is waiting on belong to the thread rather than to the tab you happen to have open. Reload in the middle of a long task and the reply carries on from where it was, with the earlier steps still on screen.

## Use Cases

### Product manager working with an AI agent on a complex deliverable

Chat with the AI employee and watch its reasoning unfold live. You see how it thinks, not just what it outputs.

### Support agent resolving multi-step customer issues in real time

The AI employee reasons through the problem, takes action, and reports back, all within the chat thread.

### Analyst using an AI agent to work through a research question

Ask, watch the agent reason step by step, and course-correct in real time if it heads the wrong direction.

### Executive getting a quick decision-support answer from their AI employee

The agent surfaces the reasoning behind its recommendation so you can trust the answer, not just accept it.

## Comparison

| Before | After |
|---|---|
| You get an answer but have no idea how the agent reached it. | Live reasoning lets you watch every step of the agent's thinking in real time. |
| Errors are only visible after the agent finishes, too late to catch. | See the reasoning as it happens and redirect before the agent goes wrong. |
| Chat feels like a black box, input in, output out. | Transparent reasoning makes collaboration feel like working with a colleague. |
| You have to trust the output without any way to verify it. | Follow the reasoning chain and validate the logic before acting on the result. |

## FAQ

### Can I stop the AI agent while it is in the middle of a response?

Yes. You can stop the agent at any point during a streaming response and redirect it with a new instruction. The agent picks up with your new direction without losing the conversation context built so far.

### What does "visible reasoning" mean in practice?

Before giving you an answer, the AI employee works through a chain of thought you can read in real time. You see how it interpreted your request, what it considered, and why it chose a particular approach.

### Does the chat support long, complex tasks or just quick questions?

Both. The chat interface handles everything from quick lookups to multi-step research and execution tasks that take several minutes. The agent maintains context across the full conversation, no matter how long it runs.

### Can I see which tools the AI employee used during a conversation?

Yes. Every tool call, including the input arguments and the result, is shown inline in the conversation. You get a full audit trail of exactly what the agent did to produce its answer.

### Can I see what my AI agent is doing while it works, not just the final answer?

Yes. The chat interface streams live reasoning and shows each tool call as it happens, so you can follow the agent's thinking in real time. There is no black box, every step is visible.

### Can I send files and screenshots to my AI employee in chat?

Yes. You can attach images, PDFs, spreadsheets, and documents to any chat message, and the employee reads them as part of your request. The reply names the file it opened as it begins, so you can see the upload landed rather than watching a blank indicator and hoping.

### Do my chat conversations get saved, or do they disappear when I close the tab?

Every conversation is saved to its thread and stays there. Close the tab, switch devices, or come back a week later and the full history is waiting, including the files you sent and the work that came out of them.

### Can I make my AI employee ask before it does something sensitive?

Yes. Any tool can be set to ask first, and the employee then pauses inside the chat with approve and reject buttons before it uses that tool. Approving picks the work back up from exactly where it stopped, and rejecting cancels that one action without ending the conversation.

> Watching the agent reason through a complex research task in real time changed how I think about AI. I could see exactly why it made each decision. That kind of transparency is rare.
>
> Leo S., Research Lead · media team

## Where Web Chat fits

Web Chat is part of Ways you talk to them.

Chat in the app with live reasoning, speak to your AI voice agent over a call, message on Slack or WhatsApp, send an email, or embed a widget on your website. Every channel connects to the same AI agent with the same memory and capabilities. Switch channels mid-conversation and your employee picks up right where you left off.

- [Ways you talk to them](/en/features/channels): Every channel. One conversation.

## Read the guide

- [Guide: Web Chat](/en/guide/channels/web-chat)

## More in Channels

- [Voice Calls](/en/features/channels/voice_channel): Voice Calls is a planned channel for speaking to an AI employee out loud and hearing it reply in real time, instead of typing. It is not available yet: there is no voice button in the workspace today, and the feature is paused while the underlying voice flow is redesigned. When it ships, a voice call will run through the same skills, tools, and memory as a typed conversation.
- [Connect an Email Provider](/en/features/channels/channel_email): Point your own SendGrid, Mailgun, or Postmark inbound-parse webhook at a tenant-scoped endpoint and every email that lands there gets turned into work for an employee. Forward support requests, sales inquiries, or internal approval emails and the employee reads the sender, subject, and body, then acts on it in the background. You get a 202 response immediately; there's no waiting on the request.
- [Personal Mailbox](/en/features/channels/channel_mailbox): Every employee gets its own email address on the shared Sistava domain the moment you hire it, no setup. Email it directly, CC it on a thread with other people, or forward something over, and it reads and replies within seconds. Free to send from, with a monthly cap by plan. Control the shared employee signature in Company Profile, or connect your own Gmail or Outlook when you want it sending as you, with no cap.
- [Send from Gmail or Outlook](/en/features/channels/connected_mailbox): The built-in employee address is free and always on, but it sends as the employee, not as you. Connect your own Gmail or Outlook instead and an employee sends and replies from your real address, in your name, with no monthly sending cap on any plan.
- [Talk to Your Employee on Video](/en/features/channels/channel_in_app_video): In-App Video is a planned channel for live, face to face video calls with an AI employee, camera and microphone, directly inside the workspace. It has no working implementation today: it appears in Settings > Technical > Channels as a Coming soon placeholder only. Once it ships, it is meant to run on the same execution pipeline as every other channel, so the employee keeps its skills, tools, training, and conversation memory on a video call. If you need an employee in a live call today, it can already join a Google Meet, Zoom, or Microsoft Teams meeting as a participant, which is a different capability and available now.
- [Chat Widget](/en/features/channels/channel_chat_widget): The Chat Widget will let you embed an AI employee directly on your public website so visitors can ask questions and get answers without leaving the page. It is on the roadmap and is not available to configure yet: today it appears in Settings > Technical > Channels as a Coming soon placeholder with no setup, embed snippet, or Manage action.
- [SMS](/en/features/channels/channel_sms): SMS is a planned channel for reaching an AI employee by ordinary text message, with no app to install. It is on the roadmap and not available today: there is no SMS number, no inbound texting, and no way to turn it on yet. Until it ships, use web chat or Telegram for mobile access from any phone.
- [WhatsApp](/en/features/channels/channel_whatsapp): WhatsApp is a planned messaging channel for talking with an AI employee from your phone. It is not connected yet: the card shows as coming soon and cannot be set up today. When it ships, it will carry the same employee memory, skills, and guardrails as web chat, Telegram, Slack, and the mailbox. Telegram and Slack are live now and give you the same access from a phone, so nothing you set up today has to be rebuilt when WhatsApp arrives.
- [Discord](/en/features/channels/channel_discord): Discord is a planned channel for bringing an AI employee into the servers and direct messages your community already uses. It will connect the same way Slack does today: authorize a bot through OAuth, then map it to one employee. It is not available to connect yet, so no Discord traffic reaches an employee until it ships. Slack, Telegram, each employee's own email address, and web chat cover the same job in the meantime.
- [Calendar Invites](/en/features/channels/channel_calendar): Add your employee's personal address as a guest on any Google, Outlook, or Apple calendar event and they show up for the meeting automatically, no RSVP or extra app on your side. Once the meeting starts, the employee joins the Meet, Zoom, or Teams call, listens, speaks when addressed, and writes a summary afterward. There is nothing to switch on for the invite itself, just turn on the Meeting Attender tool for that employee once.
- [Slack Channel](/en/features/channels/channel_slack): Bring your AI employee into Slack as a real bot user your team can mention, DM, or trigger with a slash command. Every reply lands in the same thread, and memory, tasks, skills, and tools stay shared with web chat, so a conversation started on Slack continues anywhere else. You approve or reject requests right from Slack, and the bot only speaks when mentioned, DM'd, or already in a thread.
- [Telegram Channel](/en/features/channels/channel_telegram): Connect an AI employee to its own Telegram bot so you can message it from your phone the same way you message a person. It answers direct messages instantly, and joins group chats where it replies only when mentioned or replied to. Memory, tasks, and tools are shared with the web app, so a conversation started on Telegram picks up right where it left off on the desktop, and back again.

## Explore

- [Every feature](/en/features)
- [Hire an AI employee](/en/market)
- [Pricing](/en/pricing)