When your agent runs in Agent reasoning mode, it has access to a powerful set of built-in tools. These tools let your agent search knowledge, manage files, remember users, plan complex tasks, and more — all without any setup or configuration.
Built-in tools are only available in Agent reasoning mode. Switch your agent’s reasoning mode in Settings → Reasoning.
Knowledge & Research
Your agent can search its training data and capture web pages:
| Tool | What it does |
|---|
| Search Knowledge | Searches through your agent’s uploaded documents and training data using AI-powered similarity search |
| Screenshot | Captures a visual screenshot of any web page — useful for analyzing designs, checking content, or monitoring pages |
| Web Fetch | Extracts a web page’s text content as structured markdown — useful for reading articles, documentation, or any page content |
User Memory
When User Memory is enabled, your agent can manage what it remembers about each user:
| Tool | What it does |
|---|
| Remember | Saves important details about the user (name, preferences, decisions) to long-term memory |
| Recall | Searches through saved memories to find relevant information |
| List Memories | Shows all saved memories for the current user |
| Forget | Removes a specific memory about the user |
File Operations
When your agent has access to a Workspace, it can create and manage files:
| Tool | What it does |
|---|
| Write | Creates new files or overwrites existing ones |
| Read | Reads file contents (can target specific line ranges) |
| Edit | Makes precise, targeted changes to existing files |
| Search | Finds patterns across all workspace files using regex |
| Delete | Removes files from the workspace |
Planning
For complex tasks, your agent can create structured plans and execute them step by step:
| Tool | What it does |
|---|
| Create Plan | Writes a detailed plan document and registers it for tracking |
| Update Plan | Modifies the plan, advances to the next step, or marks steps complete |
| Read Plan | Checks the current plan’s status and progress |
| Accept Plan | Submits a draft plan for your approval before execution begins |
Plans always require your approval before the agent starts executing. This gives you a chance to review, modify, or reject the approach.
Communication
Your agent can interact with users and other agents:
| Tool | What it does |
|---|
| Ask User | Presents structured questions with selectable options — great for gathering preferences or confirming approaches |
| Notify User | Sends a proactive notification when something needs attention |
| Message Agent | Sends a message to another agent in your project for cross-agent collaboration |
Code Execution
When your agent has access to a Workspace and Code Sandbox is enabled, it can run code and shell commands:
| Tool | What it does |
|---|
| Command | Runs shell commands in an isolated sandbox environment (e.g., npm install, python script.py, curl, wget). Has access to all workspace files. |
Automation
Your agent can set up tasks that run on a schedule:
| Tool | What it does |
|---|
| Schedule Task | Creates a recurring task that runs automatically (e.g., daily reports, weekly checks) |
Reasoning
Your agent has an internal thinking tool:
| Tool | What it does |
|---|
| Think | Reasons through a problem step-by-step without showing the thinking to the user — helps the agent plan its approach |
| Level | Behavior | Examples |
|---|
| Auto | Executes immediately without asking | Search, Read, Think, Recall |
| Conditional | May require approval depending on your settings | Write, Edit, Delete, Command |
| Always | Always asks for your approval first | Accept Plan, Ask User |
Best Practices
- Use Agent reasoning mode for tasks that require multiple tools (research + writing, reading + editing)
- Let your agent create plans for complex multi-step tasks — review the plan before approving
- Enable User Memory for agents with repeat users to provide personalized experiences
- The Ask User tool is great for gathering requirements before the agent starts working
- Use Schedule Task for automated monitoring, reporting, or maintenance workflows
Next Steps
