Documentation Index
Fetch the complete documentation index at: https://neuroai-cf1c4abf.mintlify.app/llms.txt
Use this file to discover all available pages before exploring further.
Overview
Tool Configuration allows you to enable, disable, and manage individual capabilities for your agents. Choose exactly which tools each agent needs, nothing more, nothing less.
This ensures:
- Focused agents — Each agent has only needed capabilities
- Security — No unnecessary access to external services
- Performance — Faster execution with fewer tools
- Cost control — Pay only for what you use
- 29 tools available — Enable what you need (soon will be more)
1. Task Management
Organize work with structured to-do lists
All 5 capabilities enabled:
Clear All
Remove all tasks and sections completely
Creates empty state for fresh start
⚠️ PARAMETER: confirm (REQUIRED)
Use when: Starting over, resetting task list
Create Tasks
Create organized task lists with sections
Features:
- Single section or batch creation
- Auto-creates sections if needed
- Order-based execution
- Edit during execution
Best for: Large, complex projects (10+ items)
Example: Research projects, multi-phase work
⚠️ PARAMETERS:
- sections (optional, array of objects)
- section_title (optional, single section)
- section_id (optional, single section)
- task_contents (optional)
CRITICAL: Pass sections as array, NOT JSON string
Delete Tasks
Remove tasks or sections
Features:
- Delete by task ID or section ID
- Section deletion removes all contained tasks
- Clean task list during execution
Use when: Tasks become unnecessary, requirements change
⚠️ PARAMETERS:
- task_ids (optional)
- section_ids (optional)
- confirm (optional, required when deleting sections)
Update Tasks
Mark tasks complete or modify content
Features:
- Mark complete immediately after finishing
- Batch multiple updates into single call
- Update content if requirements change
- Keep task list current
Best practice: Mark complete right after task finishes
⚠️ PARAMETERS:
- task_ids (REQUIRED)
- content (optional)
- status (optional)
- section_id (optional)
View Tasks
See all tasks, sections, and progress
Features:
- View current tasks and status
- Check progress regularly
- Identify next task to execute
- Verify all tasks are complete before final output
Use when: Every 2-3 task completions
⚠️ PARAMETERS: None
2. Files & Folders
Create, edit, read, and organize files
All 5 capabilities enabled:
Create File
Create new file with content
Features:
- Relative path to /workspace
- Set permissions optionally
- Supports any file type
Example: 'src/main.py' for /workspace/src/main.py
⚠️ PARAMETERS:
- file_path (REQUIRED)
- file_contents (REQUIRED)
- permissions (optional)
Delete File
⚠️ CRITICAL: Request user confirmation via ask() tool
NEVER delete without explicit user permission
Process:
1. Use ask() to request confirmation
2. Include clear message: "Do you want me to delete [file_path]?"
3. Wait for user confirmation
4. Only proceed if user confirms
⚠️ PARAMETERS:
- file_path (REQUIRED)
Edit File
Make targeted edits to existing files
Features:
- Preserves unchanged code
- Minimizes unchanged content
- Uses // … existing code … comments
- Multiple edits in single call
- Faster than full rewrite
Format edits in sequence:
// … existing code …
FIRST_EDIT
// … existing code …
SECOND_EDIT
// … existing code …
⚠️ PARAMETERS:
- target_file (REQUIRED)
- instructions (REQUIRED)
- code_edit (REQUIRED)
Full File Rewrite
Completely replace file content
Use when: Rewriting entire file (faster than many edits)
CRITICAL FOR PRESENTATIONS:
- MANDATORY for template-based presentations
- MUST preserve template structure
- Update ONLY content with research data
- Do NOT use create_slide for templates
⚠️ PARAMETERS:
- file_path (REQUIRED)
- file_contents (REQUIRED)
- permissions (optional)
Str Replace
Replace specific text in file
Features:
- Replace unique string (appears once)
- Simpler than edit_file for single changes
- Relative path to /workspace
Use when: Single, unique text replacement
⚠️ PARAMETERS:
- file_path (REQUIRED)
- old_str (REQUIRED)
- new_str (REQUIRED)
3. Web Search & Research
Find and analyze information online
2 / 2 capabilities enabled:
Web Search
Search internet for information
Features:
- Batch searches (array of queries)
- Concurrent execution
- Multiple topics simultaneously
- Results include titles, URLs, dates, direct answers
CRITICAL: Pass query as array, num_results as integer
NOT as JSON strings
Example:
- Single: "query": "AI trends 2025"
- Batch: "query": ["AI trends", "ML advances", "automation"]
Use for: Research, current info, parallel searches
⚠️ PARAMETERS:
- query (REQUIRED, string or array)
- num_results (optional, integer)
Scrape Webpage
Extract full text from multiple webpages
Features:
- Single operation extracts from multiple URLs
- Main content without HTML markup
- Optional full HTML for structure analysis
- Saves time vs. one-at-a-time scraping
Best practice: Collect multiple URLs from search results
scrape all in single call
⚠️ PARAMETERS:
- urls (REQUIRED, array)
- include_html (optional)
4. File Reader & Knowledge Base
Read and search documents
Read File
Extract text from workspace files
Supports:
- PDFs (text extraction)
- Word documents
- Text files (.txt, .csv, .json, code files)
- Images use load_image instead
Batch mode: Read up to 20 files concurrently
Much faster for multiple files!
⚠️ PARAMETERS:
- file_path (optional, single file)
- file_paths (optional, batch mode)
Use ONE, not both
Search File
Semantic search within documents
Better for large files!
Use instead of read_file for:
- Large PDFs and documents
- Finding specific information
- Multiple files with targeted search
Example: Search contract for "termination clause"
⚠️ PARAMETERS:
- file_path (for single file)
- file_paths (for multiple files)
- query (search terms)
Global KB Sync
Download knowledge base files locally
Features:
- Syncs all assigned KB files
- Creates folder structure
- Automatically searchable
- Files ready for semantic_search
⚠️ PARAMETERS: None
Semantic Search
Natural language search across files
Features:
- Searches /workspace by default
- Specify file path for targeted search
- Only searches indexed files
- Large files may take time to index
CRITICAL: Use ls_kb to check indexing status if no results
⚠️ PARAMETERS:
- queries (REQUIRED)
- path (optional)
5. Web Browser
Browse websites, click buttons, fill forms
4 / 4 capabilities enabled:
Browser Act
Perform browser actions via natural language
Features:
- Automatic screenshots with every action
- Verify displayed values match intentions
- File upload support with filePath parameter
CRITICAL FILE UPLOAD RULE:
Any action involving:
- Upload buttons
- File inputs
- Resume sections
- Choose file dialogs
MUST include filePath parameter
⚠️ PARAMETERS:
- action (REQUIRED, natural language)
- variables (optional)
- iframes (optional)
- filePath (optional, REQUIRED for upload actions)
Extract structured content from page
Uses Stagehand for intelligent extraction
⚠️ PARAMETERS:
- instruction (REQUIRED)
- iframes (optional)
Browser Navigate To
Go to specific URL
⚠️ PARAMETERS:
- url (REQUIRED)
Browser Screenshot
Take screenshot of current page
⚠️ PARAMETERS:
- name (optional)
6. Presentations
Create and manage stunning slides
8 / 8 capabilities enabled:
Create Slide
Create or update single slide
WHEN TO USE: Custom theme presentations only
WHEN TO SKIP: Not for template-based presentations
Use full_file_rewrite instead
Features:
- Parallel execution for fast completion
- 1920x1080 dimensions (16:9 aspect)
- Fixed pixel dimensions, absolute positioning
- No responsive design patterns
CRITICAL:
- Complete Phase 3 BEFORE using (research, content, images)
- Styling from custom color scheme in Phase 2
- Design for presentations, not websites
⚠️ PARAMETERS:
- presentation_name (REQUIRED)
- slide_number (REQUIRED)
- slide_title (REQUIRED)
- content (REQUIRED)
- presentation_title (optional)
❌ DO NOT USE: file_path (doesn't exist!)
Load Template Design
Reference design from presentation template
Features:
- Copy entire template to /workspace/presentations/{name}/
- Extract style patterns (colors, fonts, layouts)
- Preserve 100% of CSS styling
- Update ONLY text content
Use template as design inspiration OR edit directly
⚠️ PARAMETERS:
- presentation_name (REQUIRED)
Export Presentation
Export to PPTX and PDF simultaneously
Features:
- Both formats run in parallel
- Store locally for repeated downloads
- Or return directly
⚠️ PARAMETERS:
- store_locally (optional)
Delete Presentation, Delete Slide, List Presentations, List Slides
Management tools for presentations
Self-explanatory functions for organization
7. Image Processing
Generate, edit, and analyze images
Image Vision
View and analyze images
Understand content, identify elements
Create, edit, upscale, or remove backgrounds
Generate images and videos
Canvas Editor
Create interactive canvases for design
Add frames for export regions
AI processing (upscale, remove background)
IMPORTANT: Call ONE AT A TIME (not parallel)
Race conditions cause errors
⚠️ KEY FUNCTIONS:
- add_frame_to_canvas: Call FIRST for Instagram/TikTok sizes
Returns frame_id for use with image_edit_or_generate
- add_image_to_canvas: NEVER parallel - one at a time
- ai_process_canvas_element: Upscale or remove background
- save_canvas: Persist all elements and frame colors
8. Automation & Triggers
Set up automatic execution
7 / 7 capabilities enabled:
Create Scheduled Trigger
Schedule agent to run automatically
Features:
- Cron expressions for timing
- Template variables {{variable_name}}
- Reusable for different users
Example: 'Monitor {{company_name}} brand'
Users provide their own company names
⚠️ PARAMETERS:
- cron_expression (REQUIRED)
- prompt (REQUIRED)
- variables (optional)
Create Event Trigger
Run agent on specific events
Features:
- Event-based automation
- Compose event triggers
- Pass trigger_slug and profile_id
⚠️ PARAMETERS:
- trigger_slug (REQUIRED)
- profile_id (REQUIRED)
- trigger_config (optional)
Toggle Scheduled Trigger, Delete Scheduled Trigger, Get Scheduled Tasks
Manage scheduled automation
Enable/disable, delete, or view all scheduled runs
9. Credentials & Authentication
Manage API keys and integrations
4 / 4 capabilities enabled:
Create Credential Profile
Create authentication for toolkit
Features:
- Supports Composio toolkits
- Returns authentication link
- User completes connection
⚠️ PARAMETERS:
- toolkit (REQUIRED)
- profile_name (optional)
Add authenticated profile to agent
Use after profile is connected
Enables agent to use that service
⚠️ PARAMETERS:
- agent_id (REQUIRED)
- profile_id (REQUIRED)
Get Credential Profiles, Delete Credential Profile
View or remove existing profiles
Manage all integrations
Find information about people, companies, academic research
People Research
Find professional background information
Search and analyze people
Company Research
Find business information
Detailed company analysis
Academic Research
Search academic papers and researchers
Features:
- FREE using Semantic Scholar
- Search papers, authors, institutions
- Get paper details, citations, references
Functions:
- paper_search: Find relevant papers
- search_authors: Find researchers
- get_paper_details: Full paper information
- get_author_details: Researcher profiles
- get_author_papers: All papers by author
⚠️ PARAMETERS: Vary by function, see documentation
11. Voice Calls
Make and manage phone calls
5 / 5 capabilities enabled:
Make Phone Call
Initiate outbound call with AI agent
Features:
- AI conducts conversation
- Based on provided configuration
- Returns immediately
SAFETY: Blocks emergency numbers and illegal activity
Use: ALWAYS follow with wait_for_call_completion
⚠️ PARAMETERS:
- phone_number (REQUIRED)
- system_prompt (REQUIRED)
- configuration (optional)
Wait For Call Completion
Monitor call until complete
Features:
- Real-time conversation streaming
- Live transcription
- Continuous status checking
MANDATORY: Use immediately after make_call
Then: Use get_call_details for full transcript
Get Call Details, End Call, List Calls
Manage phone calls
View transcripts, end calls, see history
12. Knowledge Base
Store and retrieve information
9 / 9 capabilities enabled:
Global KB Upload File
Add file to knowledge base
Features:
- Upload from sandbox
- Organize in folders
- Add description
⚠️ PARAMETERS:
- sandbox_file_path (REQUIRED)
- folder_name (REQUIRED)
- description (optional)
Global KB Create Folder
Create folder for organization
⚠️ PARAMETERS:
- name (REQUIRED)
- description (optional)
List Contents, Enable Item, Delete Item
Manage knowledge base
View, enable/disable, or remove items
13. Agent Configuration
Modify agent settings
2 / 2 capabilities enabled:
Get Current Agent Config
Check current agent configuration
Use before making updates
See what's already enabled
Update Agent
Modify agent configuration
Can update:
- Name and description
- Tools and capabilities
- MCP servers
- System prompt (appended with high priority)
Call whenever user wants changes
⚠️ PARAMETERS:
- agent_id (REQUIRED)
- configuration (REQUIRED)
14. Chat & Messages
Communicate with users
Ask
Request user input or confirmation
CRITICAL for file deletion:
1. Use ask() for confirmation
2. User must explicitly approve
3. Only then proceed with deletion
Use when: User input essential to proceed
BEST PRACTICE:
- Provide follow_up_answers (2-4 clickable options)
- Keep questions concise (1-2 sentences max)
- Use natural, conversational language
⚠️ PARAMETERS:
- text (REQUIRED)
- attachments (REQUIRED when sharing results)
- follow_up_answers (optional, recommended)
✅ Do’s
✓ Enable only needed tools
Focused agents are more reliable
✓ Check current config before updating
Use get_current_agent_config first
✓ Request user confirmation for destructive actions
Always ask before deleting or major changes
✓ Use batch operations
Faster than individual calls
Multiple files, tasks, searches at once
✓ Monitor long operations
Use wait_for_call_completion, get_actor_run_status
✓ Provide clickable options
follow_up_answers when asking users
✓ Attach files when sharing results
Never share about files without attaching them
✓ Mark tasks complete immediately
After each task finishes
❌ Don’ts
✗ Don't enable unnecessary tools
Adds complexity, increases costs
✗ Don't delete without confirmation
Always ask user first
✗ Don't use old parameter names
Use EXACTLY the specified names
✗ Don't pass JSON strings for arrays
Pass actual array objects
✗ Don't run canvas operations in parallel
One at a time only (race conditions)
✗ Don't skip extraction after Apify runs
Always use get_actor_run_results
✗ Don't assume file indexing complete
Use ls_kb to check status
✗ Don't proceed without user confirmation
Especially for destructive operations
Common Configuration Scenarios
Scenario 1: Research Agent
Enabled tools:
✓ Web Search
✓ Scrape Webpage
✓ File Reader
✓ Semantic Search
✓ Task Management
✗ Browser Act (not needed)
✗ Voice Calls (not needed)
Capabilities: Search, read, analyze research
Scenario 2: Content Creation Agent
Enabled tools:
✓ Presentations
✓ Image Generation
✓ Canvas Editor
✓ Web Search
✓ File Creation
✓ Task Management
✓ Web Browser (for research)
Capabilities: Create content, presentations, images
Scenario 3: Automation Agent
Enabled tools:
✓ Scheduled Triggers
✓ Event Triggers
✓ Web Browser
✓ Web Search
✓ Task Management
✓ Credentials Manager
✓ File Creation
Capabilities: Automate workflows, scheduled execution
Scenario 4: Business Research Agent
Enabled tools:
✓ People Research
✓ Company Research
✓ Web Search
✓ File Reader
✓ Presentations
✓ Task Management
Capabilities: Research businesses and people
Configuration Workflow
Step 1: Define Agent Purpose
What will this agent do?
Example: "Research market trends and create reports"
What tools are necessary?
Example: Web Search, Research, Presentations, Files
Step 3: Get Current Config
Check what's already enabled
Use get_current_agent_config
Step 4: Update Configuration
Add needed tools
Use update_agent with new tool list
Step 5: Test Agent
Try agent with simple task
Verify all needed tools work
Adjust if necessary
Critical Parameter Rules
ALWAYS use EXACT parameter names:
- Typos cause failures
- Refer to each tool's documentation
- Example: "confirm" not "confirmation"
Array vs String:
- Queries: Pass as array [query1, query2]
- NOT as JSON string "[query1, query2]"
- num_results: Pass as integer 5
- NOT as string "5"
File paths:
- Relative to /workspace
- Example: 'src/main.py' not '/workspace/src/main.py'
- Shell commands: Use /workspace path
Batch operations:
- Pass multiple items in single call
- Faster than sequential calls
- Check tool documentation for batch support
Security & Permissions
Restrict tools by agent role:
Sales Agent:
- Web Search (research)
- Browser (company research)
- NOT: Voice Calls (if not needed)
Support Agent:
- Task Management
- File Reader
- Knowledge Base
- NOT: Presentations (if not needed)
Marketing Agent:
- Presentations
- Image Generation
- Web Search
- Social Media tools
Credential Management
Each tool may require authentication:
1. Create credential profile
2. User completes OAuth flow
3. Configure profile for agent
4. Agent can now use that service
Example: Slack, Gmail, GitHub integrations
Troubleshooting
”Parameter not found"
Check exact parameter name
Typos cause failures
Refer to tool documentation
"Array expected, string provided"
Don't pass JSON strings for arrays
Pass actual array objects
Example: ["query1", "query2"] not "['query1', 'query2']"
"File not found"
Check file path
Relative to /workspace
Use /workspace/ prefix in shell commands
Add tool to agent configuration
Use get_current_agent_config
Use update_agent to enable
"File still indexing”
Large files take time to index
Use ls_kb to check status
Try again in a moment
Getting Started
1. Choose Your Agent Type
Sales, Marketing, Research, Support, etc.
What will agent actually need?
Use update_agent to add tools
4. Test Configuration
Try agent with sample task
Verify all tools work
5. Deploy
Agent ready for production use
FAQ
Can I enable all tools?
Technically yes, but not recommended. Focused agents are more reliable and cost-effective, but Super Agent from NeuroAI has all tools on.
Can I change tools later?
Yes. Use update_agent anytime to add/remove tools.
Do I pay for unused tools?
Depends on the tool. Some charge per use, others are free. Check pricing.
How do I test a tool?
Create test task that uses that tool. Monitor for errors or unexpected behavior.
Can tool access be restricted?
Yes, through credential profiles and configuration. Some tools require authentication.
What if a tool fails?
Check error message. Usually parameter or authentication issue. Verify configuration.
Can multiple agents share credentials?
Yes. Create credential profile once, configure for multiple agents.
How do I remove a tool?
Use update_agent to remove from configuration.