Validation as a tool call.
A Model Context Protocol server that exposes DimeADozen's validation primitives to AI agents, IDE extensions, and platform partners. Score an idea, surface failure-mode patterns from comparable historical cases, and pull a structured decision document — directly from a tool call inside the conversation where the idea was raised.
Public beta launches post-Product Hunt (Tue June 9). npm package, Anthropic + Smithery directories, Cloudflare + AWS Marketplace listings.
What the server exposes.
Three primitives. Each callable from any MCP-aware client — Claude Desktop, Continue, Zed, the@anthropic-ai/sdkagent loop, or your own.
tool: validate_idea
Takes a one-paragraph business idea. Returns a 0-100 viability score, the named primary failure-mode cluster, and a structured rationale grounded in comp-set retention math from public SEC filings. Same engine that powers the free idea-score on the web.
$0.10 per call after 100/mo free tier.
tool: surface_pattern
Takes a structural attribute (category / business-model / capital-intensity / network-effect class) and returns the named historical autopsy that most-closely matches — Juicero / Munchery / Quibi / Theranos / etc — with the specific failure modes that apply to your idea.
$0.05 per call after 100/mo free tier.
tool: pull_report
Given a session ID from validate_idea, fetches the full 200+ page validation report — comp-set, retention math, regulatory drag, capex floor, go/no-go read. Streams structured JSON suitable for downstream agent reasoning.
$0.20 per call after 100/mo free tier.
Pricing.
Usage-based for builders. Flat-rate for platforms. No seat licenses, no per-user counts.
Free tier.
100 calls per month per organization across all three tools. No card on file. Suitable for individual builders and personal projects.
Usage tier.
After the free tier: $0.10 / validate_idea · $0.05 / surface_pattern · $0.20 / pull_report. Billed monthly. Per-organization rate limits at 10 RPS / 1M calls per month.
Partner contracts — $5K-$25K annual.
For accelerators, founder-coaching platforms, IDE/agent vendors, and AI-first incubators with predictable cohort volume. Flat-rate annual contract covers unlimited calls inside your organization's scope, plus a white-label option (custom subdomain + brand-aligned response framing) for Tier 1 partners.
Tiered as $5K (Tier 3) / $10K (Tier 2) / $25K (Tier 1) based on cohort size + integration depth. Higher commitments negotiated case-by-case.
Marketplace listings.
Available in every MCP catalog founders are likely to discover the server from. Same package, same free tier, same usage rate across all channels.
- Anthropic MCP Directory — featured under business-validation tooling at public-beta launch.
- Smithery — searchable as
@dimeadozen/mcp. - npm — install via
npm install @dimeadozen/mcpand add to your MCP client config. - Cloudflare Workers AI — deployable as a remote MCP endpoint with bring-your-own-domain support.
- AWS Marketplace — annual partner subscriptions available through AWS billing for enterprise procurement.
This is for you if
You're building an AI agent or IDE extension that helps founders decide what to build, and you want grounded validation primitives — citations, comp-sets, failure-mode taxonomies — instead of LLM-paraphrased generic advice. Or you're a platform serving founders and want validation primitives wired into your cohort onboarding flow.
This isn't for you if
You need raw LLM completions or a generic business-plan generator. The MCP server returns structured research-backed primitives, not free-form text generation — the value is in the underlying validation methodology + source-of-truth comp-set data, not in the surface format.
FAQ
- When does the server actually ship?
- Public beta launches the week of June 16, 2026 — directly after the DimeADozen Product Hunt launch (June 9). Builder waitlist signups get the npm package + spec + a complimentary 1,000-call beta allotment.
- Does the free tier really cover real usage?
- Yes. 100 calls per month per organization is generous enough for individual builders prototyping agents and for early-stage indie founders integrating validation into their own workflow. The pricing only matters once you cross 100 calls/mo — at which point you're running enough volume that $0.10/call is dramatically cheaper than building the underlying validation pipeline yourself.
- Is there a white-label option for partner platforms?
- Yes, available at the Tier 1 partner contract ($25K annual). Custom subdomain (e.g.
yc.dimeadozen.ai/mcp), brand-aligned response framing, and cohort-private rate-limit isolation. Tier 2/3 contracts use the standard endpoint. - How is the data sourced?
- The server returns the same comp-set + retention math that powers DimeADozen's paid validation reports — SEC filings, 10-K disclosures, named comparable companies, retention-curve math. Each tool response includes the underlying citations, so downstream agent reasoning can audit the sources.
- What about rate limits and SLAs?
- Free tier: 10 RPS, no SLA. Usage tier: 10 RPS, best-effort uptime. Partner contracts include negotiated rate limits, custom SLA, and direct support channel.
- Does the MCP server compete with my AI agent?
- No — the server exposes primitives, not a competing agent. Your agent reasons over the citations/comp-sets/failure-mode taxonomies the MCP server returns. Think of it as the validation-data layer underneath whatever agent UX you're building, not a replacement for it.
Whether you're wiring validation into an agent loop or running a founder-cohort platform that needs the underlying data — there's a path. Builders join the waitlist; platforms talk to a partner.
Questions? Email hunter@dimeadozen.ai.
We’ve been mentioned in the press
