Core Blocks
◆Block
Thinking Block
Add explicit chain-of-thought reasoning before decisions
Thinking Block
The Thinking block adds a dedicated reasoning step to your workflow. It prompts an LLM to think step-by-step before producing an answer — improving quality on complex tasks like math, logic, planning, and analysis.
Overview
| Property | Value |
|---|---|
| Type | thinking |
| Category | Core Block |
| Color | #8B5CF6 (Violet) |
When to Use
- Complex reasoning tasks that benefit from step-by-step thinking
- Math problems, logic puzzles, or multi-step analysis
- Planning and strategy tasks
- Debugging reasoning errors in your workflow
Configuration
| Setting | Type | Description |
|---|---|---|
| Thinking Prompt | Long text | What to reason about |
| Context | Long text | Data for reasoning: {{agent.content}} |
| Model | Dropdown | LLM model (o1/o3 models have built-in thinking) |
| API Key | Password | Provider API key |
Outputs
| Field | Type | Description |
|---|---|---|
content | string | Final reasoned answer |
thinking | string | Step-by-step reasoning trace |
confidence | number | Reasoning confidence score |
Example: Complex Decision Making
Goal: Analyze a business proposal with explicit reasoning before deciding.
Workflow:
[Starter] → [Thinking] → [Agent: Decision] → [Response]Configuration:
- Thinking Prompt:
Analyze this business proposal step by step. Consider: 1. Market viability 2. Financial projections 3. Competitive landscape 4. Risks and mitigation Then provide your overall assessment. - Context:
{{starter.input}}
Agent uses: {{thinking.content}} as pre-analyzed input, along with {{thinking.thinking}} for transparency.
Tips
- Use for complex tasks only — simple tasks don't benefit from explicit thinking and just add latency
- o1/o3 models have built-in reasoning — you may not need this block with those models
- Expose the reasoning trace (
{{thinking.thinking}}) in responses for transparency