Changelog

Every release, on the record.

The same standard we hold your content to: what shipped, when, and what it changed. Every date is the day the change landed in the codebase.

September 2026

Sep 1
Agent access

The door opens to agents

Trovance is operable by agents. A discovery document at app.trovance.ai/.well-known/mcp.json lists every tool the MCP endpoint serves, derived from the same registry the endpoint enforces.

  • Discovery document: app.trovance.ai/.well-known/mcp.json
  • MCP endpoint at /api/mcp/mcp, authorized by a bearer API key minted under Settings → Agent access
  • Twenty-one tools: ten reads across visibility, answers, the queue, traffic, the authority map and the Brain; one brief suggestion; ten operations, each gated by the key’s scope and revocable at any time
  • Machine notes for agents at app.trovance.ai/llms.txt

August 2026

Aug 29
Home

Home becomes a conversation

The app now opens on a question. Ask what the engines say about your market, in plain language, and the questions you asked stay in your history.

  • A chat flow on Home, with your questions kept as history
  • Instrumented, so we can see which questions people bring and answer them better
  • The greeting and the first session were rebuilt around the same thread

August 2026

Aug 15
Measurement

Agentic Web Health, published

The score became an index with a methodology you can read. How each check is scored is public at app.trovance.ai/methodology/agentic-web-health, so a number is never the whole story.

  • The methodology page: every check, and what it reads on your site
  • On-page fixes as recommendations: page-fix cards, end to end (Aug 12)
  • A lead is definable, not just a purchase: define the conversions your pixel counts (Aug 17)

July 2026

Jul 21
Coverage

All five engines

Coverage caught up with the market. Weekly sweeps grew from two engines to five: ChatGPT and Perplexity on Starter, with Google AI Overviews, Gemini and Claude added on Growth and Agency.

  • Google AI Overviews, read through SERP data (Jul 20)
  • Gemini enabled on the weekly sweep, with a per-plan engine cap (Jul 21)
  • Claude memory lane on every plan: how you’re represented across remembered context
  • Share of answer per engine, so one engine’s blind spot can’t hide

July 2026

Jul 20
Plans

Agency plan

Run the full loop for a roster of client brands from one login. Five brands are included, each with its own Brain, queue and map; more are $99 each.

  • 5 brands included, +$99 per additional brand (brand slots on Stripe, Jul 20)
  • Everything in Growth, per brand: 100 tracked questions, 10 competitors, 5 standing missions
  • 10 seats, one login for every client brand
  • Client-ready CSV exports

July 2026

Jul 15
Proof

Real-Time Analytics

One ledger for every signal. Search Console, Analytics, Bing and the AI engines themselves, merged so “how are we doing” has exactly one answer.

  • Search Console and GA4 connected read-only, with Bing search performance alongside them
  • Share of answer tracked per engine, per question, over time
  • Conversions you price yourself and Stripe payments, recorded by referral source (Jul 15)
  • The pixel’s last 30 days become queue cards on the weekly cycle (Jul 20)
The product dashboard keeps observed visibility and current work in view.
Product view captured September 2026. The product dashboard keeps observed visibility and current work in view.

July 2026

Jul 14
Proof

The live pixel

Proof moved on-site. A cookieless, first-party pixel counts human and AI visits to your pages and tells agents from humans without fingerprinting anyone.

  • Cookieless, first-party, one script tag (Jul 12)
  • AI crawlers named by the middleware snippet, so agent visits and human visits are told apart
  • Live visitor toasts and the activity feed: watch it happen, as it happens (Jul 14)
  • Agent journeys: crawl events become bot sessions on Traffic (Jul 18)

July 2026

Jul 12
Agent

Standing missions and the first agent door

Point the engine at a goal and it keeps the queue warm. Missions landed on July 3; on July 12 the MCP server let outside agents read the engine and make one safe write, behind an API key.

  • Standing missions: 1 on Starter, 5 on Growth and Agency
  • Research-first agent drafting, structured and sourced
  • The MCP server: seven reads and one safe write, keyed and revocable (Jul 12)
  • Every action leaves a receipt a human can audit: the activity ledger (Jul 14)

July 2026

Jul 10
Map

Topical Authority Mapping

The market becomes a map. The content graph was redrawn as an authority map: your tracked questions charted as clusters, with the ground you hold, the ground rivals hold, and the open territory between.

  • The authority map: clusters, routes and flags across your market
  • Pillars are subject territories, four to eight per brand, never funnel stages
  • Your company, your market and your rivals assessed on every cluster, from observed answers
The Authority Map groups questions and shows published coverage.
Product view captured September 2026. The Authority Map groups questions and shows published coverage.

July 2026

Jul 10
Production

Proactive Content Systems

From findings to a production line. Observed answers become a ranked queue of content worth making, drafted in your voice, approved by a human, published with receipts.

  • The ranked queue: impact-ordered, evidence-backed, and “no action” is a valid result
  • Author voices, trained on your edits
  • Governed publishing to WordPress and signed webhooks, receipts on every push (connectors, Jul 20)
  • Verify: the next weekly sweep re-observes the question and compares the answer against the record
A recommendation with its evidence and an explicit Accept & draft decision.
Product view captured September 2026. A recommendation with its evidence and an explicit Accept & draft decision.

July 2026

Jul 3
Brain

Compounding Intelligence

The Brain: what the brand knows, kept in one place. Everything Trovance observes becomes a node with edges, and nothing is discarded between cycles.

  • The knowledge graph: claims, pages, rivals and questions as one map
  • Six orbits kept current: brand, market, competitors, buyers, actions, results
  • Voice learned from your edits, claims kept with their proof
  • Nothing is discarded between cycles, and you can delete anything

July 2026

Jul 2
First release

The instrument ships

The first build did one thing honestly: it showed you how AI explains and recommends your brand, and refused to reduce it to a score.

  • Observed answers across ChatGPT and Perplexity, weekly
  • Evidence diagnosis: why an answer broke, and what would change it
  • The human approval gate: nothing ships itself
  • Brand intelligence captured once: voice, ICPs, competitors
  • The weekly digest, the calendar, and the first Brain (Jul 10)

Ship the next entry with us.

Every release above exists because a customer needed it.

Get started for free Read the field notes