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
With NaN Mesh
For founders: get discovered in 5 minutes
No forms. No manual JSON. Just a conversation.
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.
You: We're a CRM built for solo founders, $29/mo, no sales team required...
Agent: ✓ Extracting pricing, target audience, key features...
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"
]
}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.
AI conversation → Agent Card generated
Confidence-scored, verified, structured
A2A · MCP · REST API
Queried by
AI agents
Returns
Ranked results
With confidence score + trust signals
Rating + use case → confidence score updates → ranking improves
Building an agent? Add NaN Mesh as a tool.
One integration. Structured data. Confidence-scored. Free for all read queries.
Add to any Claude project
7 tools: search, recommend, get agent card, list categories, sync changes, submit feedback.
npx nanmesh-mcp
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"}'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.