Anthropic AI (Claude)

Send a message to Claude AI and receive a response. This is the primary way to interact with Claude for conversations, analysis, coding help, and any AI assistance. Supports streaming for real-time responses, vision for image analysis, and function calling for complex workflows. Always use this tool when you need AI assistance or to process user requests through Claude. IMPORTANT: If the user specifies a particular model (like "use Claude Opus" or "I need the most capable model"), include the model parameter to override the default.

Stream a response from Claude in real-time with live token delivery. Use this for long responses, interactive conversations, or when you want immediate feedback as Claude generates text. Supports error recovery and partial response capture if the stream is interrupted. This is ideal for user-facing applications where immediate response is important.

Stream a response from Claude in real-time with live token delivery (text only). Returns only the message text without events or metadata. Use this for direct tool response bypass mode or when you need clean streaming text output. This tool provides the same real-time streaming as stream_message but returns just the text content.

Upload a file to Claude for analysis, processing, or reference in conversations. Supports documents (PDF, TXT, CSV), images (JPG, PNG, GIF, WEBP), and other file types. Once uploaded, files can be referenced in messages for analysis, summarization, or Q&A. Files are processed securely and temporarily stored.

List all files uploaded to your Claude account. Shows file IDs, names, sizes, upload dates, and purposes. Use this to see what files are available for analysis or to reference in conversations. Files can be filtered by purpose (like assistants).

Get detailed information about a specific uploaded file including metadata, size, upload date, and status. Use this to verify file details before referencing it in conversations or to check if a file upload was successful.

Permanently delete a file from your Claude account. This action cannot be undone. Use this to clean up files that are no longer needed or to manage storage space. The file will no longer be available for use in conversations.

Analyze images using Claude's vision capabilities. Can describe image content, read text from images (OCR), answer questions about visual elements, identify objects, analyze charts/graphs, and extract information. Supports multiple image formats. Use this for any visual analysis tasks or when users upload images for processing.

Execute Python code in a secure sandbox environment. Use this for data analysis, calculations, file processing, or any computational tasks. The sandbox includes common libraries like pandas, numpy, matplotlib, requests. Code execution is isolated and safe. Perfect for processing user data, generating reports, or performing complex calculations that require programming.

Process and analyze documents (PDF, TXT, CSV, etc.) for content extraction, summarization, Q&A, or data analysis. Can extract text, tables, metadata, and answer questions about document content. Ideal for document understanding, content analysis, and information extraction tasks.

Search for information and retrieve it with proper citations and source references. Use this when you need to find specific information from knowledge bases, documents, or external sources. Results include detailed source citations for transparency and verifiability. Essential for factual accuracy and when current/specific information is needed beyond training data.

Generate structured data output in specific formats (JSON, CSV, XML, etc.) based on input requirements. Use this when you need to create formatted data, export information in specific structures, or generate data that matches exact schemas. Perfect for API responses, data exports, or structured information creation.

Interact with computer desktop through screenshots, clicking, typing, and navigation. This is a beta feature that allows Claude to see the screen, click on elements, type text, and navigate interfaces. Use for automating tasks, testing applications, or interacting with software. IMPORTANT: This requires special beta access and computer use permissions.

Send a message to Claude with custom function calling capabilities. Define specific functions that Claude can call during the conversation to create complex, interactive workflows. Use this for multi-step processes, integrations with external systems, or when you need Claude to use specific tools or APIs during the response.

Enable fine-grained streaming for tool use parameters without buffering or JSON validation. This beta feature allows Claude to stream tool parameters in real-time as they are generated, providing immediate feedback for tool calls. Use this for interactive tool usage or when you need to see tool parameters as they develop.

Enable interleaved thinking mode where Claude can think between tool calls, showing its reasoning process. This beta feature provides transparency into Claude's decision-making by revealing internal thoughts and reasoning steps. Use this for complex problem-solving, debugging, or when you want to understand Claude's thought process.

Create a message with Claude Citations API. Pass documents that Claude will cite from when answering. Claude will ground responses in the provided documents and return citations with exact quotes and character positions. Use this for RAG applications, knowledge bases, and when you need verifiable, source-backed responses.

Stream a response with Citations API. Like create_message_with_citation but streams the response in real-time. Claude will cite from provided documents and return citations at the end. Ideal for long responses where you need immediate feedback plus citation support.