Skip to main content

​
Built-in Tools

Neural Factory includes several built-in tools that agents can use out of the box. These tools are available to all agents and can be enabled or disabled per agent in the configuration. The default tool for every agent. When enabled, the agent searches connected knowledge bases to find relevant document chunks for answering questions. Results include citations back to the source documents. Allows the agent to search the internet for current information. Useful when:
  • The question is about recent events not in your knowledge base
  • The agent needs to supplement KB results with public information
  • The user asks about something outside your organization’s documents
Depending on how web search is set up for your deployment, the agent can also open a specific page when it has the link, for example one you paste into the chat or one it found in a search, and read that page instead of relying on search snippets.

​
Code execution

Runs Python code in a secure, sandboxed environment (Docker container or Azure Dynamic Sessions). The agent can:
  • Perform calculations and data analysis
  • Generate charts and visualizations
  • Process and transform data
  • Execute custom logic
Code execution is completely isolated — it cannot access your organization’s systems or data outside the sandbox.

​
Bash

Executes shell commands in the same sandboxed environment as code execution. Useful for:
  • File manipulation tasks
  • Running command-line tools
  • System-level operations within the sandbox

​
Browser

Gives the agent a web browser. Two tools in the agent’s configuration turn it on:
  • Browser control — the agent drives the browser step by step: it opens a page, reads the page’s text and its links, buttons and fields, clicks, types, presses keys, scrolls, switches tabs, and goes back.
  • Browser automation — the agent hands a whole task to a browser assistant (“find the cheapest direct flight to Lisbon on the 14th”), which works through the steps on its own and returns the result. Workflows use this tool for a browser step.
On deployments where the browser runs in the agent’s sandbox, the same sandbox that runs its code:
  • You can watch it. The Browser tab of the Agent Workspace (and the run page of a workflow) shows the page live as the agent works.
  • You can take over. Select Take control to use the page yourself, for example to sign in, pass a CAPTCHA or a two-factor check, or put the agent back on track. Select Hand back when you are done. While you hold the browser the agent cannot act in it.
  • The agent can ask you. The agent never enters passwords, pays or solves a CAPTCHA itself. When a page needs that, it hands the browser to you: the task pauses, the chat says what it needs, and nothing runs while it waits. Take control, do what is asked, then select Hand back and continue, or simply send a chat message, which resumes the task with your message as a note. In a workflow, the request appears on the run page and is answered like an approval.
  • Files meet. A file the page downloads is saved in the session and appears in the chat’s files; the agent’s code can read it in the next step. The agent can also upload a file from the session into a form on the page.
  • Its network is the sandbox’s. The browser cannot reach your organization’s internal services, private addresses or cloud metadata endpoints; your deployment’s outbound rules apply to it.
Your keystrokes and mouse movements are never recorded; that you took and handed back control is, as part of the task’s record.
Where the browser still runs on the platform’s servers, Browser control and Browser automation work as before, without the live view or takeover; the Browser tab says the deployment has no live view.

​
File editor

Reads and modifies files within the sandbox environment. Commonly used alongside code execution for tasks that involve creating or editing files.

​
Image generation

Generates images from text descriptions using AI image generation models. Useful for creating diagrams, illustrations, or visual content during conversations.

​
Memory

Allows the agent to store and recall information across conversations. The agent can:
  • Save important facts or preferences mentioned by the user
  • Recall stored information in future conversations
  • Build up knowledge about recurring topics or user preferences
Memory is scoped to the user and agent — different users have separate memory stores.