> For the complete documentation index, see [llms.txt](https://docs.diaflow.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.diaflow.io/workflow-builder/nodes/private-ai-llm-models/anthropic.md).
# Anthropic
### 1. The "Ready-to-Use" AI (Overview)
Zero setup, zero API keys. Just drag, drop, and start prompting using Diaflow's built-in AI credits.
Anthropic’s superpower is dependable high-end reasoning with live workflow support. It can handle complex business thinking, review images, remember context, and even guide next-step actions inside more advanced flows.
### 2. How to Add the Node
On your canvas, open the node picker, search for **Anthropic**, and click it to add the node to your workflow.
### 3. What can you automate? (The "Wow" Use Cases)
#### 1. **The Executive Decision Brief**
Send in long updates, policy drafts, or strategy notes and let Claude turn them into concise recommendations, trade-offs, and next actions for leaders.
#### 2. **The Visual Review Desk**
Pass screenshots, product images, or scanned visuals into Claude and get fast issue spotting, compliance checks, and plain-English feedback.
#### 3. **The Guided Operations Agent**
Use Claude in multi-step workflows where it reviews context, remembers earlier turns, and helps route the next business action in a controlled flow.
### 4. Step-by-Step: Configuration & Smart Data Injection
#### 1. **Choose your Brain (Model):**
Select your preferred model from the dropdown.
Use **Haiku** models for fast, lower-cost tasks. Use **Sonnet** for balanced speed and reasoning. Use **Opus** for the deepest analysis and highest-stakes work.
#### 2. **Write the Prompt & Inject Data:**
The easiest way to add live workflow data is to type `@` on your keyboard. Diaflow opens a visual dropdown so you can pick dynamic data from previous steps.
Typing `{{` also works perfectly if you prefer the classic bracket syntax.
Put your task, rules, tone, and desired format directly into the main Prompt box.
If you want stronger control over behavior, you can also add system instructions for Claude 3 and Claude 4 models.
#### 3. **Attach Media (Images only):**
If your workflow includes images, pass image links from earlier steps into the image field.
Claude 3 and Claude 4 models can review common image types, and you can pass more than one image when needed.
If you turn on **Memory**, Claude can keep track of earlier turns. If you turn on **Caching**, repeated requests can run faster and more efficiently.
### 5. The Output: Seeing Your Results
This node usually returns generated text that you can send straight into the next step.
Use `{{node_name.data}}` in your next step, like Gmail or Google Sheets, to pass the generated text or media forward.
In more advanced workflows, Claude can also return an action-ready result that helps the flow decide what to do next.
You may also see the model used and the final prompt that ran. This helps you review what the node actually used.
You will also see `input_token` and `output_token`. Since Diaflow handles the billing, these tokens simply act as a meter so you can track how many credits your workflow is consuming.
### 6. Golden Rules & Guardrails (CRITICAL PRO-TIPS)
#### **File/Image Limits:**
This node supports image input on Claude 3 and Claude 4 models.
Supported image types include `jpg`, `jpeg`, `png`, `gif`, and `webp`.
Very large images may be compressed automatically.
If an image is still too large after processing, or uses an unsupported format, it may be skipped silently.
For the safest results, keep images under **3.75 MB** and under **8,000 × 8,000 px**.
#### **The System Prompt Trap:**
System instructions work on current Claude 3 and Claude 4 models.
For the most reliable result, keep your most important persona, policy, and formatting instructions clear and consistent between the system field and the main Prompt box.
#### **Auto-Streaming:**
Diaflow can stream Claude answers in real time, which makes long responses feel faster in live workflow experiences.
If your step is set up for live streaming, you will see the answer appear progressively instead of waiting for the full result at the end.
#### **Streaming and advanced actions do not mix in one step:**
If you want Claude to guide the next action inside the same step, keep that step separate from live streaming.
Use one step for live response delivery and another for advanced action handling when needed.
#### **Older Claude names may be upgraded automatically:**
If you choose an older Claude model name, Diaflow may quietly run a newer Claude 4.5 version instead.
If the output style changes after switching models, re-test your workflow once before publishing.
#### **Advanced setting tip:**
On Claude 4 models, if you adjust both creativity controls at the same time, one of them may be ignored.
If results feel inconsistent, simplify the settings and focus on a clearer prompt.
#### **If a run fails:**
If you see `prompt invalid`, review the dynamic values you inserted and make sure the earlier step exists.
If you see `Invalid Model`, reselect the model and test again.
If a live response times out, reduce very large image-heavy inputs and retry.
### 7. Need Help?
* [**Understanding Diaflow Variables (`@` vs `{{ }}`)**](/agent-builder/agent-builder/cross-features-control/mention-tool.md)
* [**Contact Diaflow Support**](mailto:support@diaflow.io)
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.
Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:
```
GET https://docs.diaflow.io/workflow-builder/nodes/private-ai-llm-models/anthropic.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.