Skip to main content
Version: SunnySeal
⚠️ Alpha Release
Agent Mode is currently in Alpha.
It may be unstable and should not be used in production environments.

Smarter Reasoning Through Multi-Tool Chaining

Until now, each user query could trigger only one tool at a time, either automatically or manually.
With Agent Mode, the assistant can now plan, sequence, and combine multiple tools to achieve a specific goal.
For example, the agent can:
  1. Retrieve information from your connected documents,
  2. Complement it with real-time web search,
  3. Run a code snippet for analysis,
  4. And finally, interpret a chart or image.
All in one seamless reasoning flow.

Why It Matters

Modern business tasks rarely fit into one action.
With Agent Mode, your assistant can combine multiple sources, analyses, and reasoning steps to deliver richer, more accurate, and context-aware answers.
Example:
“Summarize this report, check the latest data online, and plot the trend.”
→ The agent will search your documents, query the web, run the code, and generate the graph — automatically.

Tools Available in the Alpha Release

This first Alpha version of Agent Mode supports the following tools:
ToolDescription
Doc SearchRetrieves information from your connected knowledge base or documents
Web SearchFetches up-to-date information from the internet
No ToolGenerates a response directly without tool invocation
Code ExecutionExecutes Python code for analysis, simulation, or computation
Visual / Image AnalysisAnalyzes and interprets visual or image content
🧪 Additional tools will be introduced in upcoming releases.

How to Enable Agent Mode

Activating Agent Mode takes just a few steps:
  1. Enable the feature in the administration console (admin access required).
  2. Open your chat settings, check “Agent Mode enabled”, and select the tools you want to make available.
  3. Refresh your interface to apply the changes.
Once enabled, users will have access to the new multi-tool reasoning experience within their chat sessions.

New Admin Dashboard Section

The introduction of Agent Mode comes with a dedicated admin section, designed to help teams monitor, debug, and optimize the agent’s behavior:
  • 🩺 Answer Flag: Investigate reported issues or incorrect responses, with access to detailed session logs.
  • ⚙️ MCP Server: Create and manage MCP servers that extend the agent’s capabilities (see related documentation).
  • 💬 Threads: View all chat sessions conducted with the agent.
    Note: In this Alpha release, session data is not yet anonymized to simplify debugging.
  • 🧰 Tools: Build, edit, or manage the tools available to the agent.

⚠️ Important Notes

⚠️ Non-Compatible Sessions
Sessions created in Agent Mode are not compatible with sessions from the standard chat mode.
Depending on which mode is active, the interface will display only the corresponding sessions.
I