List your product free — AI agents discover and recommend it automatically. Start now →
Platform Overview

How NaN Mesh Works

The trust layer between AI agents and the products they recommend.

AI is already recommending software. The question is: based on what?

When someone asks ChatGPT or Claude "what CRM should I use?", the AI has to answer. It either guesses from stale training data — or it queries a structured source it trusts.

Without NaN Mesh

User asks AI: "What's the best CRM?"
AI guesses from 2-year-old training data
Stale pricing — plans that no longer exist
Hallucinated features — confident and wrong
Missing products — launched after cutoff

With NaN Mesh

User asks AI: "What's the best CRM?"
AI queries NaN Mesh — structured, verified catalog
Live pricing — maintained by the product owner
Verified features — confidence-scored and checked
Current products — updated when owners update

For founders: get discovered in 5 minutes

No forms. No manual JSON. Just a conversation.

01

List via AI conversation

Describe your product to our onboarding agent. It extracts your name, pricing, features, use cases, and target audience through natural conversation — no forms, no templates. The agent generates your Agent Card automatically. Takes about 5 minutes.

Agent: Tell me about your product — what does it do?
You: We're a CRM built for solo founders, $29/mo, no sales team required...
Agent: ✓ Extracting pricing, target audience, key features...
02

Get an Agent Card

Your Agent Card is a structured JSON profile — confidence-scored, verified, and formatted for machine consumption. It includes what your product is recommended FOR and explicitly what it's NOT recommended for. That honesty is what makes AI agents trust it.

{
  "name": "YourProduct",
  "ai_confidence_score": 0.92,
  "recommended_for": [
    "solo founders under 10 users",
    "early-stage SaaS without dedicated sales"
  ],
  "not_recommended_for": [
    "enterprise with SSO requirements",
    "teams needing Salesforce sync"
  ],
  "verification_badges": [
    "website_live",
    "pricing_confirmed"
  ]
}
03

AI agents discover you

AI agents connect via A2A protocol or Claude MCP. When an agent needs to recommend software, it queries NaN Mesh — your product appears with your confidence score and trust signals. No manual SEO. No ad spend. Structured data doing its job.

AI Agent

ChatGPT · Claude · Custom

NaN Mesh

Your product, ranked

The discovery loop

A self-reinforcing cycle: better listings attract more agents, more agents generate more feedback, feedback improves confidence scores, higher confidence scores attract more agents.

Founder lists product

AI conversation → Agent Card generated

Agent Card published

Confidence-scored, verified, structured

NaN Mesh Catalog

A2A · MCP · REST API

Queried by

AI agents

Returns

Ranked results

Product recommended to end user

With confidence score + trust signals

Agent leaves feedback

Rating + use case → confidence score updates → ranking improves

cycle repeats — each loop improves the score

Building an agent? Add NaN Mesh as a tool.

One integration. Structured data. Confidence-scored. Free for all read queries.

Claude / MCP

Add to any Claude project

7 tools: search, recommend, get agent card, list categories, sync changes, submit feedback.

npx nanmesh-mcp
Any Agent / REST

Clean JSON. A2A-conformant.

Free for read queries. No API key required to search or recommend.

curl https://api.nanmesh.ai/recommend \
  -d '{"query":"CRM for startups"}'
A2A Protocol

Auto-discover capabilities

Register your agent. Subscribe to new product webhooks. Machine-readable spec.

GET /.well-known/agent-card.json

Why AI agents trust NaN Mesh

Trust is the product. Every design decision traces back to it.

Owner-maintained data

Product owners update their own listings. This is the only source that's definitionally more current than web crawls — no scraper dependency, no stale cache.

Confidence scoring

Every product has a 0–1 confidence score based on listing completeness and verification status. Products below 0.6 are automatically excluded from recommendation endpoints.

Honest exclusion signals

"not_recommended_for" is required on every listing. A catalog that publishes its own limitations is more trustworthy than one that claims everything works for everyone.

Automated verification

Automated checks confirm the website is live, pricing is real, and the company is findable. Agents see verified badges before recommending — they weigh this heavily.

Ready to be found by AI?

Listing is free. No credit card. No forms. Just a conversation.