🚀 Agentic Tooling: Beyond AI Workflows
Navigate to Settings > Agents > Hire Agent and toggle on "Agentic tooling" in the Optional Features section to unlock this game-changing capability!
Agentic tooling represents a fundamental leap beyond traditional AI workflows, offering unprecedented flexibility and scale in task automation and problem-solving.
🧚 Agentic Tooling 2.0 (Beta) — Now Available
Agentic Tooling 2.0 is currently in Beta and is designed to supersede the existing Agentic Tooling at a future date. It works only with reasoning models — advanced AI models that can think through problems step-by-step before acting. Using it with non-reasoning models will result in degraded performance. For full details, see Agentic Tooling 2.0.
Three Agent Modalities
When you configure an Operator agent, you can choose from three modalities that determine how the agent processes and executes tasks:
| Modality | Description | Best For |
|---|---|---|
| Operator (default) | Standard agent behaviour — responds directly to requests without autonomous tool use | Simple Q&A, straightforward conversations, quick lookups |
| Agentic Tooling (Legacy) | The original agentic tooling with a hardcoded task pipeline — the agent follows a fixed sequence of planning, execution, and verification steps | Reliable, predictable multi-step tasks; backwards compatibility |
| Agentic Tooling 2.0 (Beta) | An adaptive digital worker — autonomously decides which tools to use, in what order, and how many times, adapting its strategy as it goes | Complex, long-running tasks requiring real-time decision-making; maximum flexibility |
Agentic Tooling 2.0 relies on models that can think deeply before acting — they work through problems step-by-step and remember their reasoning as they move from one tool to the next. This deep thinking is what enables the agent to make smart decisions about which tools to use. Non-reasoning models skip this thinking step and cannot effectively drive the autonomous tool selection loop.
You can switch between modalities at any time in your agent settings under Settings > Agents > [Your Agent] > Optional Features.
For a deep dive into Agentic Tooling 2.0's architecture, capabilities, and configuration, see the dedicated Agentic Tooling 2.0 page.
🚀 What is Agentic Tooling?
Agentic tooling empowers AI agents with a comprehensive suite of capabilities that enable them to autonomously plan, reason, execute, review, and complete short to medium-term tasks. Unlike rigid workflows, agents operate using only:
- Natural language instructions - No coding or technical setup required
- Granted tools and permissions - Flexible access to capabilities as needed
🛠️ Core Capabilities
When you enable Agentic Tooling in your agent settings, you unlock:
- 🌐 Internet Search: Real-time access to current information and research
- 📚 RAG (Retrieval Augmented Generation): Intelligent document search and context retrieval from your Knowledge Hub
- 📄 Doc-by-Doc Analysis: Systematic examination of multiple documents for comprehensive understanding
- 🧠 Deep Thinking: Advanced reasoning and problem decomposition
- 💻 Code Execution: Dynamic programming and data analysis capabilities
- 🖌 Canvas: Collaborative document editing with real-time AI assistance
- 🌐 Browser: Interact with web applications as a human user would, including form filling, navigation, and dynamic content handling (Business & Enterprise with VM only)
- 📧 Email Integration: Send automated emails with dynamic content and attachments
- 📱 SMS Messaging: Send text messages for notifications and alerts
- 💬 WhatsApp Integration: Communicate via WhatsApp for broader reach
- 📅 Task Scheduling: Plan and execute future tasks autonomously
Find these options under Settings > Agents > [Your Operator Agent] > Agentic tooling
To use Email, SMS, and WhatsApp capabilities, you must first enable and configure these Agent Channels in your agent settings. Each communication channel requires proper setup before agents can utilize them.
🎛️ Execution Modes: Plan Mode vs Task Mode
When Agentic Tooling is enabled, you can choose how your agent approaches tasks through two distinct execution modes. This gives you control over the balance between speed and precision.
⚡ Task Mode (Default)
Task Mode is the default behaviour for agents with Agentic Tooling enabled. In this mode, the agent immediately begins executing tasks based on your request.
How it works:
- Agent receives your request and starts working right away
- Tools are invoked as needed without preliminary planning
- Results are delivered as quickly as possible
- Best for straightforward, well-defined tasks
Ideal for:
- Quick lookups and searches
- Simple document generation
- Routine tasks with clear requirements
- Time-sensitive requests where speed matters
🧭 Plan Mode
Plan Mode takes a more deliberate approach. Instead of diving straight into execution, the agent first creates a comprehensive plan and may ask clarifying questions before taking action.
How it works:
- Agent analyses your request thoroughly before acting
- Creates a structured plan outlining the steps it will take
- Asks clarifying questions if requirements are ambiguous
- Presents the plan for your review before execution
- Proceeds only after confirming understanding
Ideal for:
- Complex, multi-step projects
- Tasks with potential ambiguity
- High-stakes operations where accuracy is critical
- Situations requiring specific approaches or constraints
When to Use Each Mode
| Scenario | Recommended Mode | Why |
|---|---|---|
| "Find the latest sales figures" | Task Mode | Clear, simple request |
| "Create a comprehensive market analysis report" | Plan Mode | Complex, multi-faceted task |
| "Send a quick email to John" | Task Mode | Straightforward execution |
| "Develop a content strategy for Q1" | Plan Mode | Requires understanding goals and constraints |
| "What's the weather today?" | Task Mode | Simple lookup |
| "Help me restructure our pricing model" | Plan Mode | Needs clarification on objectives |
Example: The Difference in Action
Task Mode Approach:
User: "Create a presentation about our product"
Agent immediately:
- Searches knowledge base for product information
- Generates slides based on available data
- Delivers presentation draft
Fast delivery, but may miss specific requirements
Plan Mode Approach:
User: "Create a presentation about our product"
Agent first asks:
- "Who is the target audience—investors, customers, or internal team?"
- "What aspects should I emphasise—features, pricing, or case studies?"
- "How many slides would you like, and what format?"
Then presents a plan:
1. Research product documentation
2. Identify key selling points for [specified audience]
3. Create [X] slides focusing on [specified aspects]
4. Include relevant case studies and metrics
User approves → Agent executes with precision
Takes longer initially, but delivers exactly what you need
Switching Between Modes
You can switch between modes at any time:
- Navigate to Settings > Agents
- Select your agent
- Under Agentic Tooling settings, choose your preferred Execution Mode
You can also instruct the agent to switch modes conversationally. Simply say "Let's plan this out first" to engage Plan Mode behaviour, or "Just go ahead and do it" for Task Mode execution.
Best Practices
- Start with Task Mode for most everyday interactions—it's fast and efficient
- Switch to Plan Mode when the stakes are high or requirements are complex
- Use Plan Mode when you're not entirely sure what you want—the clarifying questions help refine your thinking
- Combine modes within a session: start with Plan Mode to establish the approach, then let the agent execute in Task Mode
Why Agentic Tooling Surpasses Traditional Workflows
Traditional AI Workflows: Limited by Design
- Fixed sequences: Pre-defined steps that cannot adapt to unexpected scenarios
- Rigid logic: If-then structures that break when encountering edge cases
- Manual intervention: Require human input when facing novel situations
- Limited scope: Can only handle anticipated use cases
Agentic Tooling: Dynamic and Adaptive
- Self-directed planning: Agents create their own task lists based on goals
- Contextual adaptation: Adjust approach based on real-time findings
- Error recovery: Automatically handle obstacles and find alternative solutions
- Scalable complexity: Can tackle multi-faceted problems without pre-programming
Real-World Advantages
Flexibility at Scale
While a workflow might handle 10 predefined scenarios, an agent with agentic tooling can handle thousands of variations by:
- Dynamically choosing which tools to use
- Determining the optimal sequence of actions
- Adapting when initial approaches fail
- Learning from each step to improve the next
Example Comparison
Traditional Workflow Approach:
1. Search for customer data →
2. Analyze sentiment →
3. Generate report
Fails if data format changes or new analysis is needed
Agentic Tooling Approach:
Goal: "Understand customer satisfaction trends"
Agent autonomously:
- Searches multiple data sources
- Identifies relevant patterns
- Cross-references with external market data
- Generates insights beyond predefined metrics
- Adapts analysis based on findings
The Paradigm Shift
Agentic tooling represents a shift from:
- Automation → Autonomy
- Execution → Problem-solving
- Predefined paths → Goal-oriented exploration
- Technical complexity → Natural language simplicity
This isn't just an improvement—it's a fundamental reimagining of how AI systems can work for us, breaking free from the constraints of traditional programming and workflow design.
🎯 Getting Started
Enable Agentic Tooling
You can enable agentic tooling when creating a new agent or by updating an existing one.
🆕 For New Agents
- Navigate to Settings > Agents > Hire Agent.
- Select Operator mode.
- In the Optional Features section, toggle ON "Agentic tooling."
🛠️ For Existing Agents
- Navigate to Settings > Agents.
- Select the agent you want to update.
- Scroll down to the Optional Features section and toggle ON "Agentic tooling."
We highly recommend also enabling Code Execution and Internet Search for maximum capability.
🎨 Vision Models (Available with Agentic Tooling)
When Agentic tooling is enabled, you can select from these vision models for image understanding:
- Sorcerer vision (ToothFairyAI exclusive)
- Llama 4 Maverick vision
- Llama 4 Scout vision
- Llama 3.2 11b vision
- Llama 3.2 90b vision
- Qwen 2 72b vision
💻 Agentic Coding Models
For complex agentic coding capabilities, you can select from all tool calling models:
- Mystica models (Strongly recommended for complex coding tasks)
- All other tool-calling-capable models in your workspace
💡 Pro Tips
- Deep Knowledge Search is automatically enabled with Agentic tooling for optimal results
- Code Execution becomes available only after enabling Agentic tooling
- For complex coding tasks, we strongly recommend using Mystica models for superior performance
🔍 Deep Knowledge Search: Intelligent Document Discovery
When you enable Agentic Tooling, Deep Knowledge Search automatically activates to supercharge how your AI agent finds and retrieves information from your Knowledge Hub.
Deep Knowledge Search is enabled by default with Agentic tooling—no additional configuration required!
How Deep Knowledge Search Works
Deep Knowledge Search goes far beyond simple keyword matching. It employs a multi-layered intelligence system that thinks like a researcher, not a search engine.
1. Understanding Your Intent
When you ask a question, Deep Knowledge Search first analyses your query to understand:
- What you're really looking for (not just the words you used)
- The context from your conversation history
- Your agent's specialised knowledge and instructions
This creates an optimised search strategy tailored to your specific needs.
2. Casting a Wide Net
Deep Knowledge Search doesn't rely on a single search method. It simultaneously employs multiple discovery strategies:
- Semantic Understanding - Finds documents based on meaning, not just exact words. Ask about "customer happiness" and it will find documents about "client satisfaction"
- Targeted Retrieval - Directly fetches specific documents that are most likely to contain answers
- Smart Pattern Matching - Uses AI-generated search patterns to catch relevant content that semantic search might miss
All these strategies run in parallel, ensuring comprehensive coverage while maintaining speed.
3. Intelligent Relevance Evaluation
Found documents aren't simply returned—each piece of information goes through an AI-powered relevance check:
- Is this actually answering the question asked?
- Does it fit the current conversation context?
- Is it truly useful or just tangentially related?
Only genuinely relevant information makes it through this filter.
4. Completeness Verification
Deep Knowledge Search doesn't stop when it finds some information. It continuously asks: "Have we fully answered the question?"
If gaps are identified, it automatically:
- Identifies what aspects are still missing
- Refines its search approach
- Explores alternative paths to find the missing pieces
This ensures you get comprehensive answers, not partial ones.
5. Adaptive Fallback Strategies
When initial searches don't yield sufficient results, Deep Knowledge Search intelligently escalates through additional strategies:
- Matching against your Knowledge Hub's organisational structure
- Expanding searches with related terms and synonyms
- Exploring connected topics for relevant context
The Result
Deep Knowledge Search transforms your Knowledge Hub from a static document repository into a dynamic, intelligent knowledge system that:
- Understands context and nuance - Goes beyond keywords to grasp meaning
- Finds information you didn't know existed - Discovers relevant content across your entire knowledge base
- Provides comprehensive, verified answers - Ensures completeness before responding
- Adapts its approach based on what it discovers - Learns and adjusts in real-time
For detailed settings to fine-tune Deep Knowledge Search behaviour, see the Deep Knowledge Search Settings in your agent configuration.
This is why agents with Agentic Tooling can tackle complex research questions that would be impossible with traditional search—they're not just finding documents, they're understanding your knowledge base.
🖌 Canvas: Collaborative Document Editing
Canvas is a powerful feature within Agentic Tooling that enables real-time collaboration between you and your AI agent on documents.
How Canvas Works
Think of Canvas as a shared workspace where you and your agent can work together on the same document:
You MUST create a Canvas document first - This is not optional. Click the action button at the bottom right of the Chat input section to create a new Canvas document before the agent can begin working with you.
-
Create a Canvas (REQUIRED) - You must first create a Canvas document using the UI action button
-
The agent assists - Your agent can then make intelligent adjustments to your document by:
- Creating new paragraphs
- Modifying existing content
- Deleting unnecessary sections
- Reorganizing structure
-
Real-time collaboration - Both you and the agent can make changes:
- You can edit before the agent starts working
- You can review and modify after the agent completes its changes
- The agent works recursively, refining the document based on your instructions
🧠 Knowledge Hub Integration
Canvas documents can be integrated into your Knowledge Hub, making them part of your workspace knowledge base! This means:
- Your Canvas documents become searchable resources for all agents
- They can be used in RAG (Retrieval Augmented Generation) workloads
- Build a growing library of collaborative documents that enhance your AI's understanding
Best Practices
- Minimize conflicts - Try to limit your edits while the agent is actively working on the document
- Clear instructions - Give your agent specific guidance about what changes you want
- Save to Knowledge Hub - Consider adding completed Canvas documents to your Knowledge Hub for future reference
Export Options
Once you're satisfied with your document, Canvas offers flexible export options:
- 📝 Markdown - For technical documentation and web content
- 📄 DOCX - For Word documents and formal reports
- 📑 PDF - For final, shareable versions
Canvas transforms document creation from a solitary task into a collaborative process, where AI doesn't just suggest changes but actively helps you build better content.
🌐 Browser: Web Automation at Your Fingertips
The Browser tool enables your AI agents to interact with web applications just as a human user would, opening up powerful automation possibilities for web-based tasks.
Browser capabilities are available only for Business and Enterprise subscriptions that have purchased a dedicated Virtual Machine (VM). This tool requires VM infrastructure running on your infrastructure to ensure privacy and security of your data.
What is Browser Automation?
Unlike traditional API-based integrations, the Browser tool allows agents to:
- Navigate websites - Follow links, click buttons, and interact with web elements
- Fill forms - Automatically complete and submit web forms with data
- Handle authentication - Login to password-protected websites and maintain sessions
- Extract data - Scrape information from websites, leveraging built-in search tools
- Work with dynamic content - Interact with JavaScript-heavy websites and AJAX requests
📅 Autonomous Task Scheduling: Beyond Reminders
One of the most powerful aspects of Agentic Tooling is the ability for agents to schedule future tasks autonomously, transforming how we think about long-term planning and execution.
From Reminders to Autonomous Planning
While traditional reminder systems simply notify you to do something, agentic task scheduling enables agents to:
- Self-instruct for future execution - Agents can leave detailed instructions for themselves
- Plan multi-step workflows - Schedule complex sequences of actions across time
- Adapt based on future context - Make decisions when the scheduled time arrives
Task scheduling through Jobs is available for Business and Enterprise subscriptions.
Example: Beyond Simple Email Reminders
Traditional Reminder:
"Remind me to send a project update email in 2 weeks"
Result: You get a notification in 2 weeks to manually compose and send an email
Agentic Task Scheduling:
"In 2 weeks, analyze the current project status across all active tasks,
review recent team communications, identify key accomplishments and
blockers, then compose and send a comprehensive project update email
to stakeholders with actionable insights."
Result: The agent autonomously executes this entire workflow in 2 weeks, making real-time decisions based on current data
Real-World Applications
Long-term Business Intelligence:
- Schedule quarterly market analysis reports that adapt to current trends
- Plan annual compliance reviews that adjust scope based on regulatory changes
- Set up dynamic competitor monitoring that evolves with market conditions
Adaptive Content Strategies:
- Schedule content creation that responds to trending topics
- Plan marketing campaigns that adjust messaging based on current events
- Create educational content series that builds on previous engagement
Proactive System Management:
- Schedule performance optimization reviews that adapt to current system load
- Plan security audits that focus on recently identified threat vectors
- Set up maintenance tasks that adjust based on current system health
The Paradigm Shift: From Reactive to Proactive AI
Traditional scheduling systems require you to define exactly what happens when. Agentic task scheduling enables:
- Context-aware execution - Decisions made based on conditions at execution time
- Self-modifying workflows - Agents can adjust their own future tasks based on results
- Intelligent priority management - Agents can reschedule or modify tasks based on changing circumstances
- Continuous learning loops - Each execution informs and improves future scheduled tasks
This represents a fundamental shift from "do this at this time" to "achieve this goal by this time using whatever approach works best given current conditions."
Managing Scheduled Tasks
All tasks scheduled by your agents can be reviewed, monitored, and managed through the Jobs section in your settings:
- Review scheduled tasks - See all upcoming agent-scheduled activities
- Monitor execution results - Track success rates and outputs in External chats
- Modify schedules - Adjust timing or disable tasks as needed
- Delete obsolete tasks - Clean up completed or unnecessary scheduled activities
This central management ensures you maintain full visibility and control over your agents' autonomous planning activities.
With agentic tooling, you're not just automating tasks—you're empowering AI to think, adapt, and deliver results that would be impossible with traditional workflows. Experience the future of AI assistance today!