Capture.kit

Team tier · Multi-agent plugin

Three color teams. 30 minutes per pass.

Drop a 200-page proposal draft. Get Pink, Red, and Gold review reports plus a rolled-up priority list — in customer-branded Word, in 30 minutes per color. Your humans review the judgment calls; the agent handles consistency, acronym audits, readability, and win-theme reinforcement. $3,500/yr per seat. 60-day pilot.

Buy — $3,500/yr/seatWatch the 9-minute walkthrough

  • Manual color-team reviewers

    judgment + rote

    130–200 hr × $200 / RFP

  • VisibleThread

    readabilitycompliance L/M only

    Enterprise quote

  • GovDash / Procurement Sciences

    L/M shred

    Enterprise quote

  • Capture.kit Color Team Agent

    Pink, Red, Gold passesconsistencyreadabilityacronymwin-theme

    $3,500/yr/seat

The wedge: your reviewers spend their hours on judgment, not typo-hunting

A real Red Team review on a 200-page proposal runs 130–200 reviewer-hours. Most of that time goes to consistency checking, acronym audits, readability scoring, and win-theme reinforcement counts — work that doesn't require senior judgment but absolutely has to get done before submit.

This Agent does that work in 30 minutes per color pass. Your reviewers walk in to a draft that's already been through a structural-gap audit (Pink, 65–75% completion stage), a Section M scoring proxy (Red, 85–95% stage), and a final-polish pass (Gold, 48–72 hours pre-submit). They spend their hours arguing with the agent's findings and making the judgment calls only humans can make.

The agent does not auto-edit. It surfaces findings, cites verbatim source text, and ranks by impact × effort. Every finding has a [VERIFY] posture — the human reviewer is still the gate.

ROI math

RFPs / year
6
Reviewer rate
$200/hr
Rote hours / RFP
150

Annual savings

$180,000

Break-even

1 RFP per seat per year

At 6 RFPs/yr × 150 rote hours × $200 blended rate, the agent saves a single seat $180K/yr in reviewer time. Break-even is one RFP.

How it works

  1. Drop the proposal draft into Claude Code or Cowork

    The plugin loads inside your own Claude environment. Word, PDF, and Markdown drafts all work. The orchestrator reads the draft, the SOO/RFP if you provide it, and any internal style guide.

  2. Pink Team — completeness pass (target stage: 65–75%)

    Structural gap matrix vs. Section L/M. Win-theme presence audit. Recommended sections to write next. Voice profile: collegial, focused on what's missing.

  3. Red Team — persuasiveness pass (target stage: 85–95%)

    Per-section scoring proxy 1–5 against each Section M factor. Persuasion gaps. Competitor implicit-comparison flags. Voice profile: skeptical evaluator. The hardest pass to fake — most tools skip it.

  4. Gold Team — final polish (target stage: 48–72 hr pre-submit)

    Consistency report (numbers, names, dates, claims). Fog/Flesch readability per section. Acronym audit + glossary diff. Win-theme reinforcement count. Voice profile: meticulous senior editor.

  5. Rolled-up priority list

    All findings from all three teams ranked by impact × effort. Customer-branded Word output. Drop into your existing review process — the agent reports findings, your team makes calls.

The orchestrator routes inputs to all three sub-agents (or any single color on demand) and assembles four output documents. Sub-agents share four sub-skills — consistency, readability, acronym, and win-theme — that are also sold standalone in the bundle. None of the sub-agents auto-edit; they all return structured findings with verbatim source citations.

What you get

  • pink-team-report.docx

    Customer-branded. Structural gaps, win-theme presence, recommended sections to write next.

  • red-team-report.docx

    Per-section M-factor scoring proxy, persuasion gaps, competitor implicit-comparison flags.

  • gold-team-report.docx

    Consistency, readability, acronym audit, win-theme reinforcement count, final polish recommendations.

  • rolled-up-priorities.docx

    All findings ranked by impact × effort, with `[VERIFY]` flags on inferred elements. The single document the proposal manager actually walks through.

Also included

  • plugin.json — Cowork / Claude Code plugin manifest
  • agents/ — orchestrator + three color-team sub-agents
  • skills/ — consistency, readability, acronym, win-theme (each sold standalone in Wave 2)
  • templates/ — branded .docx report templates
  • benchmarks/ — three reference proposals with human-Red-Team findings ground truth
  • admin-dashboard/ — read-only HTML page for seat usage and per-RFP runs
  • INSTALL.md, README.md, ETHICS.md, SECURITY.md, PRICING.md, UPGRADE.md, CHANGELOG.md
  • onboarding/kickoff-loom-script.md + 30-min-walkthrough-agenda.md

See it run on a real proposal

9:14200-page services proposal. All three color passes. Rolled-up priority list. Recorded against a synthesized proposal built from public SAM.gov solicitations and APMP sample winners.

Want the actual benchmark numbers? Read the 3-RFP recall report — services, SaaS, and construction proposals, agent findings vs. human-Red-Team ground truth, no email gate.

Pricing

$3,500per seat / year
  • Rolling updates included for the duration of the subscription
  • 60-day pilot — full refund if the agent doesn't outperform a manual reviewer pass on your first real proposal
  • NET-30 invoicing available for 5+ seats

The 3-seat tipping point. Three seats run $10,500/yr. The Proposal Stack Bundle is $8,500/yr for 10 seats and includes the Compliance Matrix Skill, the Acronym Builder Skill, and the L/M Crosswalk Prompt Pack on top. If you have 3+ reviewers, the bundle is the rational buy.

How we handle your data

Public-data-context-only positioning

The agent runs inside your own Claude environment. Proposal content never leaves your account. The AUP forbids CUI, SSI, and source-selection-sensitive material — that's a hard line, not a recommendation.

Zero retention via Anthropic API contract

Anthropic's zero-retention API contract applies to all agent runs. The plugin does not centralize, cache, or transmit proposal content beyond your Claude session. SECURITY.md covers this in plain English for your InfoSec team.

Not FedRAMP authorized

Stated plainly because it matters: this product is not FedRAMP authorized. For FedRAMP-required workflows, use AWS Bedrock GovCloud with a Claude variant — that's a separate product on the roadmap. Don't deploy this on classified or controlled-unclassified workflows.

Verify before submit

Every finding cites verbatim source text and section reference. Inferred elements — page allocation suggestions, owner assignments, edge-case classifications — are tagged [VERIFY] in every report. False-confidence outputs are the worst-case failure mode in this product, and we engineered against them: findings without a verifiable source string get flagged for the user, not silently included.

≥80% recall vs. human Red Team

Benchmark across 3 reference proposals (services, SaaS, construction) shows ≥80% recall against a senior human reviewer's findings. The remaining gap is what the human review still owns — and the agent is honest about it.

Per-run API cost ≤ $25

Costs land in your own Anthropic account. Prompt caching on color-typology references, chunked streaming for >150-page drafts, and Sonnet-by-default with Opus only on Gold final pass keep a 200-page proposal under $25 per full run.

Admin dashboard, no central data store

Each subscription includes a read-only admin dashboard showing seat usage and per-RFP run counts — labeled by your admin, never proposal content. The dashboard is a magic-link URL tied to your subscription. No SaaS account, no central database holding your work.

Who buys this

If you're a Capture VP, BD VP, or Proposal Director at a mid-tier prime — or a principal at a proposal shop running 4–6 color teams per major proposal with 4–6 human reviewers per team — this is for you. Your reviewers are paid $150–$250/hour blended. They spend 130–200 hours per RFP on rote audit work that doesn't require their judgment. This agent compresses that work into 30 minutes per color pass.

If you bid 1–2 RFPs a year, the Compliance Matrix Builder Skill at $499 is a better fit. If you have a department of 5+ reviewers, the Proposal Stack Bundle at $8,500/yr/10 seats is cheaper per seat and adds the Compliance Matrix and the L/M Crosswalk pack.

Who shouldn't buy this

Frequently asked

How does this differ from VisibleThread, GovDash, or Procurement Sciences?

Three differences. One: scope. Those tools shred Sections L and M for compliance. We run the actual color-team passes — Pink for completeness, Red for persuasiveness, Gold for polish. Two: pricing. $3,500/yr/seat on a personal-card-or-PO threshold, vs. enterprise quote. Three: posture. The agent runs in your own Claude environment with zero retention. No central proposal database, no SaaS scope creep.

We have a security questionnaire — can you respond?

Yes. SECURITY.md is a 1-page plain-English summary built for InfoSec review. It covers: zero retention, public-data-context-only positioning, no central data store, subprocessor list (Stripe for payments, Resend for email, Cloudflare for downloads, Anthropic for the agent runtime). The full questionnaire response template ships with the plugin and includes language for the typical CMMC, NIST 800-171, and ISO 27001 questions. We are explicit that we are not FedRAMP authorized.

How does seat licensing work?

License key + seat count baked into the plugin config. Each reviewer activates a seat with a machine fingerprint. The admin dashboard tracks seat allocation but does not gate runs — gate failures cost more than over-allocation does. If a reviewer leaves, the admin reassigns the seat from the dashboard via Stripe Customer Portal.

What's the per-run API cost on our Anthropic account?

≤ $25 for a 200-page proposal across all three color passes. Sonnet by default, Opus only on the Gold final pass for the harder edge cases. Prompt caching on the color-typology and APMP references keeps cost well below the price ceiling. We document the exact cost profile in PRICING.md.

What's the 60-day pilot?

Buy a seat, run the agent on at least one real proposal in the first 60 days. If the agent doesn't outperform a manual reviewer pass on consistency, acronym, readability, or win-theme reinforcement — full refund, no questions. The bar is honest, not exhaustive.

Does this replace our color-team reviewers?

No. It compresses the rote 130–200-hour audit work into 30 minutes per color pass. Your reviewers still own the judgment calls — what to fix, what to push back on, what to ship. The agent is explicit at the top of every report: this does not replace a human Gold Team. Verify before submit.

Can we customize the report templates with our branding?

Yes. The .docx templates are editable Word files in templates/. Drop in your logo, fonts, and section structures. The agent renders into your template, not ours.

What about NET-30 invoicing and multi-seat purchase orders?

Available for 5+ seats. Stripe Customer Portal handles the billing infrastructure; for a custom PO arrangement, the receipt email includes a contact path for procurement. We do not require accounts or SSO; the seat license is the gate.

Does the agent learn from our past proposals?

No. The agent is stateless across runs by design — partly for security posture (no central proposal database), partly because most shops don't want last quarter's losing proposal influencing this quarter's pursuit. If you want the agent to weight against your win themes, paste them into the run input. The Win Theme Reinforcement Skill (also in the bundle) handles this explicitly.

Bundle

Proposal Stack

$8,500/yr

10 seats

What's in it

  • Color Team Reviewer Agent (this product) — 10 seats
  • Compliance Matrix Builder Skill — handles Sections L, M, C, and H
  • Acronym Builder Skill
  • L/M Crosswalk Prompt Pack
  • Win Theme Reinforcement Skill

Three Color Team seats run $10,500/yr standalone. The bundle is $8,500 with 10 seats and adds the Compliance Matrix Skill plus three more proposal-shop tools. The break-even is 3 reviewers — almost every shop that runs formal color-team review crosses that threshold.

See the Proposal Stack — $8,500/yr/10 seats

Ready to give your reviewers their hours back?

Buy — $3,500/yr/seatRead the 3-RFP benchmark

60-day pilot. NET-30 for 5+ seats. Runs in your own Claude environment.