Anthropic Claude

Add Claude to a flow: messages, structured replies, streaming, tools, classification, token counts, memory and batches.

Guides

Blocks

Add these blocks on the canvas by searching for their name. Each block page lists its pins and how it behaves.

• create batch anthropicconnection_batch_create Submit many prompts as one asynchronous batch at 50% of the standard price (results within 24h). Wire on_finished/on_item to be called automatically when the batch completes (the system polls in the background), or poll the batch_id yourself with the get batch block.
• get batch anthropicconnection_batch_get Poll a batch by id. Fires `ended` and returns all results (text/usage/cost per custom_id) once processing finishes.
• build output template anthropicconnection_build_schema Pick the fields Claude should return and describe each one; produces an output template for a Claude block's output template. Each field's type comes from the field itself; per field you set a description and (optionally) the allowed options. No JSON Schema knowledge needed.
• build tool anthropicconnection_build_tool Define a tool Claude can call and pick the function that runs it. The function receives the model's arguments as in-pins and returns the result back to Claude. Feed the output into a message block's tools pin.
• classify anthropicconnection_classify Classify input into exactly one of your labels using Claude; route the resulting label with a switch block.
• count tokens anthropicconnection_count_tokens Count the input tokens a prompt (with system + attachments) would use, and estimate the cost; before you send it. Free; no spend.
• create memory anthropicconnection_create_memory Open (or create) a persistent memory by id. The same id returns the same memory across runs, so Claude can pick a conversation back up weeks later. Feed the memory into a send message block.
• delete memory anthropicconnection_delete_memory Delete a memory by id (forget the conversation and its files).
• get memory anthropicconnection_get_memory Read a memory's contents by id (conversation history, memory files, summary) without creating one.
• get usage anthropicconnection_get_usage Report total tokens and cost used on this connection (USD and the connection's currency), with a per-category breakdown.
• list memories anthropicconnection_list_memories List all memories (id, timestamps, expiry, turn count) for browsing or housekeeping.
• memory from JSON anthropicconnection_memory_from_json Cast a raw {id, guid} object into a memory handle for a send message block; for passing a memory handle you already have.
• send message anthropicconnection_message Send a message to Claude and get the reply, with built-in cost guard, model allowlist and typed error reporting.
• stream message anthropicconnection_message_stream Send a prompt to Claude and stream the reply token-by-token to a WebSocket channel as it generates, then return the full text.
• parse JSON (repair) anthropicconnection_parse_json Recover and parse JSON from messy text (code fences, prose, trailing commas, truncation). No API call.
• preflight document anthropicconnection_preflight Convert a document (PDF, image, HTML, spreadsheet, text) to clean Markdown with a cheap model, stripping boilerplate; run before sending it to an expensive model to cut tokens.
• preflight documents anthropicconnection_preflight_batch Convert several documents to clean Markdown with a cheap model, in parallel. Returns one .md file per input plus the concatenated Markdown; run before sending an upload set to an expensive model.
• output template from JSON anthropicconnection_schema_from_json Cast a raw JSON Schema object into an output template for a message block (or a tool's parameters); for full control beyond the build output template block.
• summarize memory anthropicconnection_summarize_memory Compact a memory: fold the older turns into a running summary and trim the history, so a long conversation stays cheap and in-context while keeping its gist. Run it periodically.
• tool if anthropicconnection_tool_branch Pick a tool based on a condition: outputs the true tool when the condition holds, otherwise the false tool. Add elseif branches for more cases.
• code execution tool anthropicconnection_tool_code_execution Let Claude run Python in a sandbox (data analysis, math, charts); Anthropic-hosted. Wire the tool output into a message block's tools pin.
• tool from JSON anthropicconnection_tool_from_json Cast a raw JSON object into a tool for a message block's tools pin; for full control beyond the build tool block.
• memory tool anthropicconnection_tool_memory Let Claude manage its own long-term memory notes (read, write, search) for the wired memory. Drop into a message block's tools pin; use the toggles to limit what it may do.
• multiple anthropicconnection_tool_multiple Combine several tools into one connection for a message block's tools pin.
• tool switch anthropicconnection_tool_switch Pick a tool by matching an expression against case values (with an optional default). Wire the result into a message block's tools pin.
• web search tool anthropicconnection_tool_web_search Let Claude search the web (Anthropic-hosted). Wire the tool output into a message block's tools pin. Optionally cap uses or restrict domains.
• open connection function_create_anthropicconnection Create a connection to the Anthropic (Claude) API for use by Claude blocks.

Next steps