Chat

​

Overview

Chat is the core entry point and default page of Super Agent—the main stage where users collaborate with the AI Agent. In a unified conversation window, users submit requests in natural language, and the Agent automatically breaks down tasks, calls tools, displays its thinking process, and streams every intermediate artifact (search results, files, code, screenshots, form questions) back into the conversation flow in real time. What Chat helps you do:

• Submit tasks in natural language — Hand vague requirements to the Agent in a single sentence; the Agent decides which tools to call on its own.
• Fully observable execution — Thinking, Tool Call, AskUserQuestion, and Authorization are all visible, expandable, and interruptible in the conversation flow.
• Multi-modal input — Text, images, documents, audio/video, code snippets, and Drive files can all serve as context.
• Session persistence — Automatically saves session history, supports search, and one-click sharing as read-only links or saving as templates.
• Task orchestration entry — Schedule the current message as a recurring task directly in the input bar, without leaving Chat.

​

Target Users

Scenarios not recommended:

• Lightweight “Q&A bot” scenarios that don’t need tool calls or file outputs—a general LLM is sufficient, no need for Super Agent.
• Real-time control scenarios requiring millisecond-level precision.
• High-compliance scenarios requiring sessions and data to be fully local and never leave the enterprise network (consult the Enterprise team).

​

Use Cases

• Daily research and reporting — Describe a topic in Chat; the Agent automatically searches, synthesizes, and outputs results as downloadable Markdown / Word / Excel.
• Code and script tasks — Upload code or a repo link; have the Agent read, modify, run, and return diffs.
• Document processing — Drag in PDF / Excel / PPT directly; have the Agent summarize, extract, rewrite, or translate—with optional Office Sync for two-way sync.
• Multi-round forms / information gathering — The Agent proactively confirms key parameters via AskUserQuestion to avoid going in the wrong direction.
• Parallel multi-Agent execution — Enable multi-session mode to dispatch the same request to multiple Agents / models in parallel and compare results.
• Scheduled execution — Click the calendar icon in the input bar to turn the current message into a task that runs automatically daily / weekly.
• Session reuse — Save proven workflows via Save as Template for one-click reuse later.

​

Quick Start

Complete your first successful Chat conversation in 3 minutes.   1. Open https://mulerun.com and log in (Google account or email).   2. After login, you’ll land on the Chat page by default—session history is on the left, the input box is in the middle.   3. Click New Chat in the upper-left corner (shortcut ⌘K / Ctrl+K) to create a new session.   4. Enter your request in the input box, e.g.:   “Search for ‘2026 AI video generation model comparison’ and compile it into a Markdown report.”   5. Press Enter to send. The Agent will display, in sequence: Thinking → Tool Call → output files.   6. Once the task completes:     - Click Download on the output file to save locally.     - Click Share in the upper-right to generate a read-only sharing link.     - Click Save as Template to save this session as a template. Once complete, you should see a full report and download entry. If no output appears, check the balance hint above the input box, or refer to Section 7 (FAQ).

​

Detailed Guide

1. Create and Manage Sessions Steps:     1. Click New Chat in the upper-left (⌘K / Ctrl+K) to enter a blank session.   2. Scroll through history sessions in the left sidebar, or use Search at the top to find by keyword—matching sessions are highlighted.   3. Click More … on a session item to rename, favorite, share, or delete.   4. The session title is auto-generated by the Agent after the first message is sent—you can also manually click the title to edit. Sessions are private by default; deleted sessions enter a recovery period and become unrecoverable after expiry. 2. Input Box The input box is Chat’s “console,” centralizing attachments, mode switching, scheduling, and more.   Attachment upload:

• Click the + button on the left → choose Local File or From Drive.
• Supports images, documents, audio/video, code, archives, and other common types.
• Attached files appear as Chips above the input box, removable individually.

Slash Commands / Skills:

• Start with / to trigger the command palette.
• To install Skills, use the SkillBadge command format (see 5.6).

Mode switching:

• Single-Agent mode (default) — One Agent processes one message serially.
• Multi-Agent parallel mode — After switching, the same message triggers multiple Agents / models in parallel for comparison.

Schedule Chip (scheduled sending):

• Click the calendar icon before sending to open the Cron / natural language time editor.
• Once saved, this message is registered as a scheduled task, executed automatically on schedule (execution records in Router Console → Tasks Log).

Office Sync:

• When uploading Excel / PowerPoint, enable Sync on the Chip.
• Once enabled, Agent modifications to the file are written back to the source file in real time, and local edits sync to the Agent.

3. Message Stream The Agent’s reply consists of several atoms, each observable and expandable:

• Thinking / Thought — The Agent’s extended reasoning process (collapsible).
• Plan — A list of sub-steps broken down from the task (collapsible).
• Tool Call — Tools called by the Agent (search, browser, code execution, file operations, etc.); expand to see inputs and returns (collapsible).
• Tool Call Group — Consecutive calls of the same tool type are aggregated into one group (collapsible).
• AskUserQuestion — The Agent proactively asks you a question—single-select / multi-select / free input (not collapsible; must be answered to continue).
• Authorization Card — OAuth or API Key authorization dialog (not collapsible).
• File Result — Generated files with preview, download, and copy-path actions (not collapsible).
• Turn Comment — Rate or leave feedback on a single answer (collapsible).

Message actions:

• Hover over a message → Copy / Regenerate / Like / Dislike / Branch from here.
• Click Stop during a long-running task to interrupt the Agent.
• Any Agent question that timed out unanswered can be re-answered from history.

4. SkillBadge and Skill Installation The SkillBadge format lets you bring third-party Skills into a single message in one go: /skill-finder install skill_id directly @{owner}/{repo}#{skillName}:{version} {your actual message} Behavior:

• A gray SkillBadge label appears above the input box and the sent message bubble.
• The command prefix /skill-finder install … does not appear in the message body—only your actual message is shown.
• Multiple Skills can be stacked at once: /skill-finder install skill_id directly @owner/repo#skill-a:latest @owner/repo#skill-b:latest do something

5. Share, Export, and Templatize

• Share (session) (upper-right Share button) — Generates a read-only sharing link, optionally with an access code.
• Save as Template (upper-right more menu) — Save the current session as a template for one-click reuse via New Chat later.
• Export message content (single message hover menu → Copy / Copy as Markdown) — Copies to clipboard.
• Export output files (File Result card → Download) — Single-file download, or aggregate to Drive.

6. Edit, Pause, and Delete Editable targets and their effect:  

• Session title — Edit the name; takes effect immediately; no impact on in-progress tasks.
• Scheduled message — Edit Cron, prompt, attachments. Cron / prompt changes take effect on the next run; attachment changes take effect immediately. The currently executing instance is not interrupted.
• Share link — Edit access code, visibility, expiration; takes effect immediately; already-distributed links validate against the new policy immediately.

Actions:

• Stop (halt execution) — Use when the Agent drifts or you don’t want to wait further. Immediately interrupts the current Tool Call; outputs from this turn are preserved. Recoverable—you can continue asking questions in this session.
• Pause (scheduled task) — Use when you don’t want it to run temporarily. Preserves Cron and history; no longer triggers. Recoverable—can be re-enabled anytime.
• Delete (session) — Use when no longer needed. Session and associated files enter a recovery period. Recoverable within 30 days via Trash; irreversible after expiry.

​

Plans, Quotas, and Limits

Below is a capability comparison; refer to https://mulerun.com/pricing for specific quota numbers. General usage limits:

• Single file size and per-upload file count limits depend on the plan; split if exceeded.
• Per-message context and per-session length have limits; earlier messages are automatically compressed when exceeded.
• Sessions with long inactivity are not deleted, but related Agent state may enter cold-start.
• Share links are subject to anti-abuse policies—high-frequency access may be rate-limited.

​

FAQs

Q: What’s the difference between Chat and a regular ChatBot? A: Chat is more than just conversation—the Agent actually calls tools like search, browser, code execution, and file operations behind the scenes. Every step is visible in the conversation flow, and real files can be produced. Q: How does the Agent know what I want it to do? A: When the Agent is unsure, it proactively confirms via AskUserQuestion. Stating the goal / scope / output format clearly in your first message reduces repeated questioning. Q: Can I retract a sent message? A: Messages themselves cannot be retracted, but you can click Stop to interrupt the Agent at any time, or Regenerate / Branch from the message. Q: How are Credits consumed? A: Billed comprehensively based on the model used and tools called (search, code execution, file processing, etc.). Each Tool Call displays estimated / actual consumption on its card. Q: How large can attachments be? Can the Agent see the full content? A: Single file size depends on the plan. The Agent reads necessary excerpts first; over-long documents are automatically chunked or retrieved, not necessarily loaded into context all at once. Q: When I drag in an Excel file, do the Agent’s changes sync to my local file? A: With Office Sync enabled, changes sync bidirectionally. Without it, the Agent’s modifications produce a new file without changing your original. Q: How do I make a message run automatically every day? A: Click the calendar icon in the input box before sending and configure the time or Cron expression. Execution records are in Router Console → Tasks Log. Q: If I share a session with someone, can they continue sending messages in it? A: Share links are read-only by default—others can only view history, not continue chatting. For collaboration, invite members within the team Workspace. Q: Can deleted sessions be recovered? A: Recoverable within 30 days via Trash; irreversible after expiry. Q: Where do I report issues? A: @ MuleTeam directly in Chat, or visit https://mulerun.com/contact.