From product update to every AI agent. In minutes.

Toneeto is a verification pipeline. It takes what you say about your product, cross-references it against public evidence, and distributes the verified result to AI agents via an open protocol.

The problem: AI agents are stuck in the past

AI agents learn about products from training data, web scrapes, and retrieval systems. All of these are delayed by weeks or months. When you ship a feature, change pricing, or deprecate an endpoint, agents keep citing the old information until they happen to re-crawl your site — if they ever do.

There is no inbox for AI agents. No way to push an update. No standard for "here's what changed." SEO doesn't help — agents don't use PageRank. Ads don't help — agents don't see ads.

ChatGPT

Says your API is in beta

Reality: GA since January

Claude

Cites $29/seat pricing

Reality: You raised to $49 three weeks ago

Perplexity

Misses your Salesforce integration

Reality: Shipped last month

Change Objects: a standard format for product truth

A Change Object is a structured, versioned snapshot of everything that's true about a product right now. Pricing plans, features, integrations, audience fit, recent changes — all with sources, confidence scores, and freshness metadata.

Think of it like a schema.org record, but designed for AI agents instead of search engines. Every field is machine-readable. Every claim is traceable to evidence.

change_object / close.com

Plans: Essentials ($49/seat), Professional ($99/seat), Enterprise (custom)

verified_at: 2026-02-24 · verified_until: 2026-03-24

Event: "Raised Essentials price from $29 to $49/seat"

fact_date: 2026-02-01 · confidence: 0.91 · sources: 3

Claim: "Built-in calling and SMS available on all plans"

solidity: cross_ref · sources: pricing page, G2 review, blog post

Decision: choose if you need built-in dialer · avoid if you need marketing automation

freshness: current · trust_score: 0.87

How verification works

Every claim in a Change Object goes through a multi-source verification pipeline. We don't trust any single source — including the publisher. The pipeline gathers evidence from independent channels, cross-references them, and scores confidence.

Evidence gathering

We collect from 7+ independent sources: your pricing page (via JS-rendered fetch), social mentions (X/Twitter), YouTube reviews, ad intelligence (Google + Meta), web search, and AI analysis.

Claim extraction

Each piece of evidence is analyzed to extract specific, falsifiable claims — "Essentials plan costs $49/seat/month" not "pricing is competitive." Claims that can't cite a source are dropped.

Cross-referencing

Claims confirmed by 2+ independent sources get "cross_ref" solidity — the highest confidence tier. Single-source claims are labeled honestly. Contradictions are flagged.

Freshness scoring

Every claim gets a TTL (time-to-live) and freshness tier: breaking (hours), recent (days), current (weeks), or baseline. Expired claims are flagged and re-verified automatically.

The evidence trail: nothing is taken on faith

Every claim is labeled with its provenance. Consuming agents can see exactly how confident a claim is and why. This is the opposite of how LLMs normally work — where the model states facts with no attribution.

cross_ref

Confirmed by 2+ independent sources. Highest confidence.

Pricing verified against website + G2 review

publisher_stated

The business explicitly declared this. Trusted but single-source.

Business submitted feature list via portal

single

Found in one source only. Lower confidence, marked honestly.

Feature mentioned in one blog post

inferred

Derived from patterns in evidence. Lowest confidence tier.

Market position estimated from ad spend

For agent developers: query, don't scrape

If you're building an AI agent that recommends products, compares tools, or answers pricing questions — you need accurate data. Not training-data-era guesses. Toneeto gives you structured, verified product intelligence via MCP or REST API.

MCP Server — claude_desktop_config.json

{
  "mcpServers": {
    "toneeto": {
      "command": "uvx",
      "args": ["toneeto-mcp"]
    }
  }
}

Structured data

JSON responses with plans, features, claims, and metadata. No parsing HTML or guessing formats.

Source trails

Every claim cites sources. Your agent can tell users "verified against 3 independent sources."

Freshness metadata

TTLs, verified_at timestamps, and freshness tiers. Your agent knows how current the data is.

See it in action.

Run a free audit on your domain and see what AI agents are getting wrong — with real evidence.