RankBrain

Changelog

Server 2.10.0 · Skill 3.5.0 · Harness rules 3.5.0

We publish calm, customer-focused notes when RankBrain AI gains new capabilities or becomes more reliable. Your project connection rules can upgrade automatically on the next MCP session — you do not need to re-paste the dashboard prompt for every release.

v2.10.0skill 3.5.02026-07-15New capability

Automatic harness upgrades

RankBrain AI now keeps your project connection rules up to date on each MCP session — no need to re-paste the dashboard prompt when we ship improvements.

  • New sync_project_harness tool upgrades .grok/rules, .cursor/rules, and AGENTS.md markers automatically
  • Version stamp at .search-everywhere/version.json tracks your project forward (never downgrades)
  • Changelog page documents customer-facing updates with clear, neutral language
  • Session-start gate: MCP retries while connecting instead of falling back to local brand files
  • Short reel scripts use ads and business corpus fallback until dedicated short-reels library finishes syncing
v2.9.1skill 3.3.12026-07-15New capability

Multi-domain RankBrain AI

One API key now powers SEO, Meta Ads, Google Ads, business decisions, and short-form video scripts.

  • Five-domain dashboard with live corpus status
  • MCP-native workflow rules with attribution on every deliverable
  • generate_short_reel_script for record-ready short video scripts
v2.8.0skill 3.2.02026-07-13Improvement

Harness auto-setup

One dashboard paste configures Grok, Cursor, Claude Code, and Codex with your real API key.

  • Grok stdio transport fix for reliable authentication
  • Cloud skill document served only via MCP — not in paste prompt
  • Silent-mode audits with exact before/after fix tables

How we write updates

Professional, calm, forward-looking. Lead with what works better for the customer.

We share

  • New capabilities customers can use today
  • Reliability and connection improvements
  • Corpus or domain expansions (neutral, benefit-focused)
  • Clear upgrade path when setup steps change

We avoid

  • Internal bug blame or embarrassing incident detail
  • Competitor criticism or political commentary
  • Security vulnerability specifics (say "security hardening" only)
  • Unreleased roadmap promises or ETAs we cannot keep
  • User-shaming language ("you forgot to…")