Claude Anthropic Integration

200,000 Tokens.
The Entire Tender Document.
In One Context.

Claude Sonnet 4.6 handles a 200-page RFP, your company's full product catalog, and five past proposals — all in a single context window. No chunking required. No lost context. The model reasons over the entire document at once.

Add Claude to Your Workflow → Configuration Reference

Available Models

Three Claude Models.
One for Every Workflow Step.

Claude Sonnet

Best for Most Workflow Steps

The recommended default. Balances deep reasoning, instruction-following, and speed. Handles complex RFP documents, cross-referencing requirements, and generating compliant, well-structured proposal text.

200k token context window
Strong long-document comprehension
Reliable structured JSON output
Good cost/quality balance

Use the current Sonnet model configured in your workspace.

Claude Haiku

Best for Fast, Simple Steps

Lowest latency and lowest cost Claude model. Use for formatting steps, short classification tasks, summary generation, and any step where the input is clean and the output is formulaic.

200k token context window
Fastest response time of Claude family
Ideal for high-volume steps
Lowest cost per token

Use the current Haiku model configured in your workspace.

Claude Opus

Best for Deepest Reasoning

Highest capability Claude model. Use for the most complex analytical tasks: interpreting ambiguous contract language, multi-document contradiction detection, strategic proposal framing, and novel problem solving.

200k token context window
Best complex reasoning performance
Higher cost — use selectively
Extended thinking capability

Use the current Opus model configured in your workspace.

Best Use Cases

Where Claude Outperforms.

RFP Requirement Extraction

Feed the entire RFP document to Claude Sonnet. Ask it to extract every requirement, classify each one (technical / commercial / compliance / schedule), and identify ambiguities that need clarification. The 200k context window means no document needs to be split.

Long-Document Analysis

Contracts, technical specifications, compliance frameworks — Claude can hold hundreds of pages in context and reason across the full document. Identify contradictions, extract key clauses, summarize obligations.

Proposal Drafting

Given extracted requirements, knowledgebase context from past proposals, and a system prompt defining your tone and format, Claude Sonnet drafts proposal sections that read as though written by a senior estimator — not a template-filler.

Compliance Review

Claude checks a draft proposal against the original RFP requirements. It identifies which requirements are addressed, which are missing, and which responses are weak or non-compliant — before a human reviewer spends time on it.

Configuration

Every Setting
Explained.

When you add a Claude AI block to a workflow, you configure the following settings in the block's properties panel:

Model — select Sonnet 4.6, Haiku 4.5, or Opus 4.7
System prompt — defines the block's role and output format
Temperature — 0.0 for deterministic extraction, 0.3–0.7 for drafting
Max tokens — cap on response length
Context injection — knowledgebase, prior block output, or static text
Output schema — optional JSON schema for structured output enforcement
Fallback model — use a faster Claude model if Sonnet is rate-limited

Claude Block Config Example

{
  "provider": "anthropic",
  "model": "selected-sonnet-model",
  "system": "You are an RFP analyst for
  a construction equipment firm.
  Extract all requirements from the
  attached RFP. For each, return:
  id, category, text, compliance_risk.",
  "temperature": 0.1,
  "maxTokens": 8192,
  "context": [
    {
      "type": "knowledgebase",
      "id": "rfp-library-2024",
      "topK": 5,
      "threshold": 0.72,
      "citeSources": true
    }
  ],
  "outputSchema": {
    "requirements": [{
      "id": "string",
      "category": "string",
      "text": "string",
      "complianceRisk": "low|medium|high"
    }]
  },
  "fallbackModel": "selected-fast-fallback-model"
}

Example Workflow

Claude + Knowledgebase
for RFP Analysis.

Trigger: RFP Document Received

Email attachment or SharePoint upload triggers the workflow. The PDF text is extracted and passed as the input document.

Claude Sonnet Block: Requirement Extraction

The full RFP text (up to 200k tokens) is sent to Claude Sonnet with a system prompt instructing it to extract and classify every requirement. Knowledgebase context from past winning proposals is injected (topK=5). Claude returns a structured JSON array of requirements.

Workbook Write: Requirements Tracker

The extracted requirements array is written to a Workbook — one row per requirement, with category, compliance risk, and a draft response field populated by Claude.

Human Gate: Estimator Review

The estimating team reviews the workbook. They approve, edit, or flag rows. The gate releases when all high-compliance-risk rows are approved.

Proposal Generation + Submission

Approved workbook data is passed to a Haiku formatting block that populates the proposal template. Output is a Word document ready for the estimator's final sign-off before submission.

200,000 Tokens. The Entire Tender Document. In One Context.

Three Claude Models. One for Every Workflow Step.