AgentTools.<tool>(tool_name="...", ...) or agent.add_tool.<tool>(tool_name="...", ...). Every tool requires a unique tool_name=. Each entry lists the tool’s configuration parameters. See the Agent reference for attaching and running tools.
agent — Agent
The complete agent configuration including instructions, tools, inputs, and outputs
Select the LLM model to be used
Dynamic inputs generated from agent inputs configuration
Dynamic outputs generated from agent outputs configuration
Select the LLM provider to be used
Whether to use an existing agent
The ID of the agent to be used
code_interpreter — Code Interpreter
Execute Python code in a sandboxed environment with file I/O support. Supports spreadsheet analysis, data visualization, and generating artifacts. To read or edit uploaded files, pass them in the files input as an actual array of file references; those files are mounted at /work/inputs/{filename} (curly braces shown for clarity; replace with the exact filename). Continued sessions restore prior inputs and generated files, so work can resume even if Python memory reset. Save output files to /work/outputs/ to have them collected as artifacts; call save_file(…) for durable/versioned user-visible files. Example: files=[“$history.69dd11a49033ae4d8e65ab33”] and pd.read_csv(‘/work/inputs/data.csv’); plt.savefig(‘/work/outputs/chart.png’). You will see stdout and stderr outputs after each execution step. Include an existing session id to keep executing code within the same session.
Platform docs: Code Interpreter
The Python code to execute in the sandboxed environment
Optional files to make available to the code (e.g., spreadsheets, CSV files). Pass an actual array of file references, such as [“history.69dd11a49033ae4d8e65ab33"] or ["object.knowledge_base_item.6875a39904d4beff85bdde61”]. Files passed here are mounted into /work/inputs/<exact filename>; continued sessions also restore prior session files.
create_skill — Create Skill
Create a new skill with a name, description, and content
The full content/instructions of the skill. Use markdown formatting: headings (##, ###) to organize sections, bullet points (- or *) for instructions and guidelines, numbered lists (1. 2. 3.) for sequential steps, bold for key terms and critical instructions, and
code for technical terms or placeholders. Add a <br> tag between major sections for visual spacing.The name of the skill
A brief description of what the skill does
custom_tool — OpenAI-native built-ins (code_interpreter, file_search)
OpenAI-native built-ins (code_interpreter, file_search)
generate_chart — Generate Chart
Use this to generate a chart from. Convert a tabular file, dataframe or table to a chart or graph visualization. Supports bar, line, pie, scatter, and donut charts/graphs.
The type of chart to generate (bar, line, pie, scatter, donut)
One of:
bar, donut, line, pie, scatter, time_seriesThe dataframe to visualize as a chart. In case of csv, the input should be a properly formatted valid csv string with appropriate headers. When chaining from a tool that returned a CSV file (e.g. fetch_ratios.table), set dataframe_type=“file” and pass $action.<id>.table — never paste the upstream formatted_text/XML here.
Actual column names from the dataframe. It should be a string representing the comma separated list of column names.
The type of dataframe to be used. Only available options are table, csv, md, json, file.
The title to display on the chart
The column name to use for the X-axis (categories). Must EXACTLY match (case-sensitive) one of the column names from the CSV header.
Column name(s) for Y-axis values. Must EXACTLY match (case-sensitive) column name(s) from the CSV header. For multi-series charts (multiple lines/bars), use comma-separated names like ‘revenue,profit,cost’.
How to aggregate values when there are multiple data points per category
A brief description or subtitle explaining what the chart shows
Where to display the chart legend
The order to sort the data
The label to display on the X-axis. Defaults to the field name if not provided.
The label to display on the Y-axis
Currency symbol (only used with currency format)
Initial time period tab selection
How to format values
Comma-separated icon URLs for each series (company logos, game art, product images, etc.), in same order as y_axis_fields. Leave empty for no icons.
Comma-separated display names for each Y-axis field, in same order as y_axis_fields (e.g. ‘MSFT,NVDA’ for tickers, ‘Halo,Fortnite’ for games, ‘US,EU,APAC’ for regions).
Set to ‘true’ to show Original/% Change toggle
Scale suffix for values
get_skill — Get Skill
Retrieve the full content of a skill by its ID
The skill to retrieve. Specify by its id like $object.prompt.id
knowledge_base — Knowledge Base Reader
Semantically query a knowledge base that can contain files, scraped URLs, and data from synced integrations (e.g., Google Drive).
Platform docs: Knowledge Base Reader
Use additional LLM calls to analyze each document to improve answer correctness
The unit of retrieval. Chunks will return the most relevant chunks from the knowledge base as well as their text content. Documents will return the document metadata as well as most relevant snippets from the document. Pages will return complete pages with all chunks from pages containing relevant content
One of:
chunks, documents, pagesFormat the context for the LLM
Whether to stream the LLM response
Select an existing knowledge base, Use $object.knowledge_base.? syntax
The query will be used to search documents for relevant content semantically. Must not be empty, only include relevant information for retrieval or metadata filter generation. Generally expand any specific acronyms or abbreviations but include the original acronym or abbreviation as well
The alpha value for the retrieval. 1.0 is pure vector search and 0.0 is pure lexical search
Extract separate questions from the query and retrieve content separately for each question to improve search performance
Do a natural language metadata query
Expand query to improve semantic search
Expand query terms to improve semantic search
Use an LLM to generate metadata filters to refine your query. Agents should usually leave this false and provide filters directly in the filter input.
The score cutoff
The number of relevant chunks to be returned
Transform the query for better semantic search
The mode to use for the advanced search
The model to use for the QA
knowledge_base_agent — Knowledge Base Agent
Query a knowledge base using an agentic approach with tools.
Platform docs: Knowledge Base Agent
Optional additional context to help the agent understand the query better (e.g., conversation history, user preferences)
Select an existing knowledge base. You must provide the id in $.object.knowledge_base.id format
Select the LLM model to be used by the agent
The natural language query. The agent will use this to determine the best way to query the knowledge base. Include the key criteria needed to answer the query.
If enabled, generates a synthesized answer from the knowledge base
Select the LLM provider to be used by the agent
Controls the query effort: ‘fast’ for quick answers, ‘focused’ for balanced depth, ‘deep’ for thorough analysis
If enabled, shows an additional context input to provide context to the agent
If enabled, returns the relevant context/chunks used to generate the answer
knowledge_base_create — Create Knowledge Base
Dynamically create a Knowledge Base with configured options
Platform docs: Create Knowledge Base
To analyze document contents and enrich them when parsing
Apify API Key for scraping URLs (optional)
The overlap of the chunks to store in the knowledge base
The size of the chunks to store in the knowledge base
The name of the collection to store the knowledge base in
The embedding model to use for the knowledge base. Format: provider/model
The embedding provider to use
The file processing implementation to use for parsing documents
Whether to create a hybrid knowledge base
The name of the knowledge base to create
The precision to use for the knowledge base
The method to break text into units before chunking. ‘words’: splits by word; ‘sentences’: splits by sentence boundary; ‘paragraphs’: splits by blank line/paragraph.
Whether to shard the knowledge base
The vector database provider to use
Strategy for grouping segmented text into final chunks. ‘sentence’: groups sentences; ‘markdown’: respects Markdown structure (headers, code); ‘dynamic’: optimizes breaks for size using chosen segmentation method.
knowledge_base_fetch_document_content — Knowledge Base Fetch Document Content
Fetch the full content of a specific document from a knowledge base by scrolling through all its chunks
Platform docs: Knowledge Base Fetch Document Content
item id of the item to fetch content from
Select an existing knowledge base
Maximum number of chunks to include in the content
Number of chunks to skip for pagination.
Optional page ranges to fetch, e.g. ‘1-5,8,10-12’. If empty, fetches all pages.
knowledge_base_fetch_items — Knowledge Base Fetch Items
Advanced knowledge base item fetching with traversal, filtering, and output shaping capabilities
Platform docs: Knowledge Base Fetch Items
JSON filter object, example filter syntax: ’{“type”: “condition”, “field”: “file_extension”, “operator”: “eq”, “value”: “pdf”}’ or ’{“type”: “condition”, “field”: “title”, “operator”: “match”, “value”: “Pantheon”}’ only use filters if explicitly requested by the user. Do not use filters on fields other than file type. Available operators include are eq and neq, do not use filter operator contains
Type of items to return: FOLDERS, DOCUMENTS, or ALL
Select an existing knowledge base
Maximum number of items to return, capped at 200
Maximum depth for descendants traversal, use 0 to traverse up as deep as possible to get to {{limit}} items, use 1 to get direct children etc
Number of items to skip for pagination
Filter to list items in specific folders by providing comma-separated root IDs
Comma-separated fields to sort by
Comma-separated sort orders: asc or desc
Output verbosity level: metadata (no summaries), short (short summaries), or long (short and long summaries), Use metadata for for simply finding folders/ documents, use short for understanding folder/document contents, use long with a lower limit for deep understanding of a particular folder/subfolder
knowledge_base_get_item_bboxes — Knowledge Base Get Item Bboxes
Fetch OCR bounding boxes for specific pages of a PDF document in a knowledge base
Item ID of the document to fetch bounding boxes for
Select a knowledge base
Page ranges to fetch bounding boxes for, e.g. ‘1-5,8,12’
knowledge_base_loader — Knowledge Base Loader
Load data into an existing knowledge base.
Platform docs: Knowledge Base Loader
Select the type of data to load
One of:
File, URLThe file to be added to the selected knowledge base. Note: to convert text to file, use the Text to File node
The knowledge base to load data into
list_objects — List Objects
List VectorShift objects accessible to the current user
Include objects shared with you by other users
Maximum number of objects to return (0 = no limit)
The type of VectorShift object to list
Number of objects to skip (0 = no offset)
Return full object data instead of just id, name, and type
mcp_connector — Integration-backed MCP (Slack, GitHub, etc.)
Integration-backed MCP (Slack, GitHub, etc.)
mcp_server — Mcp Server
Bind an MCP server (created via web UI) as a tool source
pipeline — Pipeline
Platform docs: Pipeline
The ID of the pipeline to be used
read_memory — Read Memory
Load the full content of a memory entry. Use this when: - A memory shown in context appears truncated or abbreviated - A knowledge base memory needs its full document index (folder structure, document list) - You need the complete text of a memory that shows [content not loaded] Returns the full content as a string. Knowledge base memories: returns the KB’s document index and folder hierarchy (not document content). Use knowledge_base_query to search document content.
The memory_id shown in the context_memories section next to the memory name.
transformation — Transformation
Use Python code to create a custom node
Platform docs: Transformation
The ID of the transformation to be used
write_memory — Write Memory
Create or update a Memory object for use with context-aware agents. You can store text memory content and or references to object types like knowledge bases, knowledge base items, files and prompts.
Content to store in the memory. Each item can be plain text or an object reference using ‘object.knowledge_base.617’, ‘$object.file.718’). Allowed object types: KnowledgeBase, File, KnowledgeBaseItem, Prompt.
brief description of what this memory contains and when it should be used
Use this to update an existing memory instead of creating a new one.
Name of the memory entry
Optional explicit user ID to use for User scope. If empty, authenticated caller user ID is used. Provide the raw ObjectId only (e.g. ‘6875a39904d4beff85bdde61’), not prefixed with ‘user:’ or any other label.
Scopes to associate with this memory. Allowed values: User, Session, Agent, Org, Workspace. Memory will be accessible if all scopes are true. All listed scopes are required (AND). Example: Agent + Workspace means memory is only in scope for that agent inside that workspace.