Agent B

From Zero to Proposal: A Complete Guide to Using RFP2Proposal

Part 3 of 3 | A 3-Part Series on Automating Proposal Generation with AI | Part 1 | Part 2

You’ve read the pitch. You’ve seen the business case. Now let’s get your hands dirty.

This guide walks you through everything: installing the plugin, configuring it for your organization, analyzing an RFP, generating a full proposal, and exporting a polished Word document. By the end, you’ll have a working proposal pipeline.

Prerequisites

Before you begin, you’ll need:

  • Claude Desktop app with Cowork mode enabled (currently in research preview)
  • Node.js (version 18 or higher) — needed for the Word document export server
  • An RFP document — PDF or Word format. If you don’t have one handy, you can use any procurement document to test.

Step 1: Install the Plugin

Option A: Claude Cowork (Desktop App)

Open Claude Desktop and navigate to Customize > Browse Plugins > Upload. Select the rfp2proposal plugin folder. Claude will detect the plugin structure and install it.

Once installed, you’ll see the plugin’s commands available as slash commands in your conversation.

Option B: Claude Code (CLI)

If you prefer the command line:

claude plugin install --path /path/to/rfp2proposal

Verify Installation

Start a new conversation in Cowork and type:

/analyze-rfp

If Claude responds asking for an RFP document, the plugin is working. If it doesn’t recognize the command, double-check that the plugin folder structure matches what’s expected (a .claude-plugin/plugin.json file at the root, with commands/ and skills/ directories).

Step 2: Set Up Your Organization Knowledge

This is the most important step. The more your org-knowledge files are filled in, the fewer placeholders your proposals will have and the more they’ll sound like your organization wrote them.

Run the Setup Wizard

Type:

/setup-org

Claude will scan your org-knowledge/ directory and show you a completeness report:

File                     | Status      | Completeness
-------------------------|-------------|-------------
company-profile.yaml     | Found       | 0% (template only)
capabilities.yaml        | Found       | 0% (template only)
methodology.yaml         | Found       | 0% (template only)
case-studies.yaml        | Found       | 0% (template only)
team-bios.yaml           | Found       | 0% (template only)
pricing-templates.yaml   | Found       | 0% (template only)
legal-boilerplate.yaml   | Found       | 0% (template only)
differentiators.yaml     | Found       | 0% (template only)

The wizard walks through each file in priority order, asking you conversational questions and generating the YAML content from your answers.

Priority 1: Company Profile

Claude will ask for your basics: company name, address, website, elevator pitch, employee count, founding year, and core values. For government contractors, it also asks for DUNS number, CAGE code, UEI, NAICS codes, and set-aside categories.

This data feeds into the cover letter, executive summary, and compliance sections of every proposal.

Priority 2: Capabilities

Your service lines, technology stack (languages, frameworks, cloud platforms, databases), certifications (ISO 27001, SOC 2, CMMI), security clearances, and active contract vehicles.

This data drives the Technical Approach section. When an RFP requires AWS and your capabilities file lists AWS as a core platform, Claude writes your technical approach with that context built in.

Priority 3: Delivery Methodology

Your methodology name and description, standard project phases, PM framework, QA approach, risk management process, and the tools you use (Jira, Confluence, Azure DevOps, etc.).

This feeds directly into the Management Approach section. If you’re an Agile shop, Claude writes about sprint cadences and ceremonies. If you’re Waterfall, it adjusts accordingly.

Priority 4: Differentiators

This is where you define your win themes — the 3-5 things that make your firm the best choice. Each win theme needs a claim and evidence. For example:

win_themes:
  - theme: "98% team retention rate"
    evidence: "Over the past 3 years, we've maintained a 98% retention rate on government contracts, compared to the industry average of 72%."
    use_in: ["executive_summary", "staffing_plan", "past_performance"]

Win themes get threaded across multiple proposal sections — introduced in the Executive Summary, substantiated in the Technical Approach, and proven in Past Performance.

Priority 5: Case Studies

For each past project, provide: project name, client, contract type and value, period of performance, description, technologies used, team size, quantified outcomes, relevance tags, and a reference contact.

The relevance tags are key. When Claude analyzes an RFP and detects it’s for a healthcare client needing cloud migration, it automatically selects your case studies tagged with “healthcare” and “cloud-migration.”

Priority 6-8: Team Bios, Pricing Templates, Legal Boilerplate

Team bios feed the Staffing Plan. Pricing templates provide the structure for the Pricing Summary (though actual dollar amounts always remain as REQUIRED placeholders — the AI never fabricates pricing). Legal boilerplate populates the Terms & Conditions appendix.

You Can Skip and Come Back

The wizard lets you skip any file and return later. Run /setup-org again at any time — it updates existing files rather than overwriting them. Your org knowledge builds incrementally over time.

Manual Editing

Every org-knowledge file is a plain YAML file. You can edit them directly with any text editor. The wizard is convenient, but not required.

Step 3: Set Up the Word Document Export Server

The plugin includes a local MCP (Model Context Protocol) server that generates formatted .docx files. You need to install its dependencies once:

cd /path/to/rfp2proposal/servers/docx-generator
npm install

That’s it. The server starts automatically when you use /export-docx or /export-matrix. It runs locally on your machine via stdio — no cloud services, no external connections.

Step 4: Analyze Your First RFP

Now for the exciting part. Upload an RFP document to your Cowork conversation (drag and drop a PDF or Word file), then type:

/analyze-rfp

Claude reads the entire document and extracts information across 12 categories:

  • Project Overview — Issuing organization, solicitation number, contract type, set-aside status
  • Scope of Work — Deliverables, phases, activities, boundaries, period of performance
  • Technical Requirements — Mandated technologies, integrations, security, performance SLAs
  • Timeline — Proposal due date, Q&A deadlines, milestones, go-live target
  • Budget Signals — Stated budget, pricing model, cost evaluation weight, CLIN structure
  • Evaluation Criteria — Factors, weights, subfactors, evaluation method (best value vs. LPTA)
  • Compliance Requirements — HIPAA, SOX, FedRAMP, CMMC, clearance levels, insurance
  • Submission Requirements — Page limits, font, margins, volume structure, file format
  • Personnel Requirements — Key personnel positions, minimum qualifications, clearance levels
  • Past Performance — Number of references, recency, relevance criteria, format
  • Pain Points — Stated problems, legacy issues, urgency indicators, previous failures
  • Incumbent Signals — Current contractor, transition requirements, dissatisfaction signals

Each item is marked as Found, Inferred, or Not Found, with confidence levels for inferred items and source section references.

For long RFPs (50+ pages), Claude processes section-by-section: Table of Contents first, then SOW, then Evaluation Criteria, then Instructions to Offerors, then everything else. It cross-references to catch contradictions between sections.

Claude presents the extraction summary and asks you to confirm key details, flag errors, and add context the RFP doesn’t contain (like “we have an existing relationship with this agency”).

Step 5: Generate the Full Proposal

Once you’re satisfied with the analysis, run:

/generate-proposal

Or, if you’re starting fresh with a new RFP, just use /generate-proposal directly — it runs the analysis as part of the pipeline.

Claude now generates all 10 proposal sections sequentially:

Section 1: Cover Letter

A formal transmittal letter with your company details from company-profile.yaml, the solicitation reference from the RFP, and a brief value proposition. Signature block is a REQUIRED placeholder.

Section 2: Executive Summary

Opens with the client’s problem (not your company history), presents your solution overview, threads your win themes from differentiators.yaml, and provides proof points from your case studies.

Section 3: Understanding of Requirements

The RFP’s scope rewritten in your voice — not copy-pasted, but paraphrased to demonstrate comprehension. Includes key challenges identified (including between-the-lines insights) and numbered assumptions for FFP proposals.

Section 4: Technical Approach

Solution architecture based on RFP technical requirements mapped against your capabilities.yaml. Every technology recommendation cites the specific RFP requirement it satisfies. Includes a REQUIRED placeholder for an architecture diagram.

Section 5: Management Approach

Your delivery methodology from methodology.yaml, customized for the specific RFP. Sprint structure, governance and reporting cadence, QA process, risk management, and a transition-in plan if incumbent signals were detected.

Section 6: Staffing Plan

Team structure matched to RFP-required positions from Category 9. Bios pulled from team-bios.yaml and tailored to this RFP. If your bios file is empty, REQUIRED placeholders are inserted for every mandated position.

Section 7: Past Performance

Case studies selected by relevance matching — industry, technology, scale, contract type. Each formatted with challenge, solution, quantified results, technologies used, and a reference contact placeholder.

Section 8: Pricing Summary

Structure and format only. The plugin provides the pricing model recommendation based on budget signals, effort estimation framework, and cost summary table structure — but every dollar amount is a REQUIRED placeholder. AI does not fabricate pricing.

Section 9: Compliance Matrix

Auto-generated by the compliance-mapping skill. A cross-reference table mapping every RFP requirement to the proposal section where it’s addressed, with status: Compliant, Partial, or Non-Compliant.

Section 10: Appendices

Structured list of supporting documents with appropriate placeholders: resumes, certifications, terms and conditions, company registration documents, insurance certificates.

Progress and Review

Claude shows progress after each section and presents a final summary:

REQUIRED:  12 items — Must be filled before submission
SUGGESTED:  8 items — Should be filled for a stronger proposal
OPTIONAL:   4 items — Nice-to-have enhancements

COMPLIANCE STATUS:
  COMPLIANT:     34 requirements
  PARTIAL:        6 requirements (placeholders to fill)
  NON-COMPLIANT:  1 requirement (gap to address)

You can ask Claude to regenerate any section with different emphasis, add context, or restructure. The proposal is a living document in your conversation — iterate until it’s right.

Step 6: Export to Word (.docx)

When you’re satisfied with the proposal, type:

/export-docx

The docx-generator MCP server produces a professionally formatted Word document with:

  • Title page with project name, your company name, and submission date
  • Table of contents (auto-generated from heading structure)
  • Consistent headers and footers with page numbers
  • Color-coded smart placeholders so your team can visually scan for what needs attention
  • Proper fonts, margins, and spacing (matching RFP submission requirements if specified)

The file is saved to your workspace folder, ready for your team to open in Microsoft Word, fill in the placeholders, and submit.

Step 7: Export the Traceability Matrix (Optional)

For proposals that require a traceability matrix — and for your own QA process — run:

/export-matrix

This generates an Excel (.xlsx) file with:

  • RFP section-to-proposal section mapping
  • Coverage status for each requirement (Covered, Partial, Gap)
  • Color coding (green for covered, yellow for partial, red for gaps)
  • Auto-filters for quick navigation
  • A summary sheet with overall coverage statistics

Evaluators love traceability matrices. They make it easy to verify that your proposal addresses every requirement. Including one — especially when it’s not explicitly required — signals professionalism and attention to detail.

Tips for Getting the Best Results

Fill in your org-knowledge files thoroughly. The single biggest factor in proposal quality is how much organizational context Claude has. A 60%-complete org-knowledge base produces proposals with 15-20 placeholders. A 90%-complete base might have 5-8. Invest the time upfront — it pays dividends on every future proposal.

Review and correct the RFP analysis. Claude’s extraction is thorough, but not infallible. If it misidentifies the contract type or misses a requirement, correct it before generating the proposal. The analysis is the foundation — errors here propagate downstream.

Don’t skip the “add context” step. After the RFP analysis, Claude asks if you have additional context. This is where you add intelligence that isn’t in the RFP: your relationship with the client, insider knowledge about what they value, competitor intelligence, lessons from previous bids with this agency. This context dramatically improves proposal quality.

Iterate on sections, don’t start over. If the executive summary doesn’t quite hit the right tone, ask Claude to revise it with specific feedback (“make it more client-focused, less about our history”). Regenerating a single section is fast; regenerating the entire proposal is unnecessary.

Build your case study library over time. After every project delivery, add a case study to case-studies.yaml. After every win, update your differentiators. Your org-knowledge is a living asset — the more you invest in it, the better every future proposal becomes.

Plugin Architecture at a Glance

For those who want to understand what’s under the hood or plan to customize:

rfp2proposal/
  .claude-plugin/
    plugin.json              # Plugin metadata (name, version, description)
  commands/
    analyze-rfp.md           # /analyze-rfp workflow
    generate-proposal.md     # /generate-proposal pipeline
    compliance-checklist.md  # /compliance-checklist workflow
    export-docx.md           # /export-docx workflow
    export-matrix.md         # /export-matrix workflow
    setup-org.md             # /setup-org wizard
  skills/
    rfp-analysis/            # 12-category extraction framework
    proposal-writing/        # 10-section generation framework
    placeholder-classification/  # Three-tier placeholder system
    compliance-mapping/      # RFP-to-compliance checklist mapping
    industry-compliance/     # Regulatory framework detection
    traceability-mapping/    # RFP-to-proposal section mapping
  org-knowledge/
    company-profile.yaml     # Your company details
    capabilities.yaml        # Services, tech stack, certifications
    methodology.yaml         # Delivery methodology
    case-studies.yaml        # Past performance examples
    team-bios.yaml           # Key personnel profiles
    pricing-templates.yaml   # Rate cards and pricing models
    legal-boilerplate.yaml   # Standard T&C and compliance statements
    differentiators.yaml     # Win themes and competitive advantages
  servers/
    docx-generator/          # MCP server for Word doc generation
  .mcp.json                  # MCP server configuration

Every markdown file in skills/ and commands/ is human-readable and editable. Want to change how the executive summary is structured? Edit skills/proposal-writing/SKILL.md. Want to add a new section to the proposal? Add it to the framework. Want to create a skill for your specific industry vertical? Create a new folder under skills/ with a SKILL.md file.

What’s Next for RFP2Proposal

The plugin’s roadmap includes two major phases:

Phase 2: Connected Knowledge Base — Pull case studies from your document management system, auto-match team bios from HR systems, and learn from past winning proposals stored in SharePoint or Google Drive.

Phase 3: Win Intelligence — Track which proposals won vs. lost, learn winning language and pricing patterns per client type, and flag risk early (“RFPs like this have a low win rate for us”).

Contributions are welcome. The plugin is MIT-licensed and designed to be extended. Check the CONTRIBUTING.md for guidelines on adding skills, commands, and connectors.

Get Started

  1. Clone the repo: git clone https://github.com/agentbee0/rfp2proposal.git
  2. Install in Claude Cowork or Claude Code
  3. Run /setup-org to configure your org knowledge
  4. Upload an RFP and run /generate-proposal
  5. Review, iterate, and export

Your next proposal doesn’t have to start with a blank page.

RFP2Proposal is an open-source project licensed under MIT. Contributions welcome.
GitHub: github.com/agentbee0/RFP2Proposal

Leave a comment