Changelog

What's new in Crossagent.

Every release, from the first @mention to today. Active development, fast releases, public history.

Built in 3 days · Launched Oct 5, 2026

v2.4.0

October 5, 2026

Current

Cross-tab native routing, onboarding overlay, and Pro tier launch.

Routing

–Native routing now works even when the target AI is on a different tab that was never opened — extension opens it in the background and restores last conversation URL.
–Improved DOM selectors for Perplexity, Grok, Mistral, and Copilot after UI changes.
–execCommand("insertText") is now the primary injection method, with fallback cascade.

Product

–First-run onboarding overlay: guides new users through installing and using @mentions.
–Pro tier launched — unlimited memory API calls, full web dashboard, memory history search.
–Popup now shows API call count, memory count, and API cap with a usage bar.

v2.3.2

October 5, 2026

Patch

Stability fixes for Roles tab, Groq simulation, and @all synthesis.

Fixes

–Roles tab now live-refreshes when switching agents without reopening the panel.
–"Groq simulating X" note shown in panel header when API relay is used.
–@all synthesis now correctly uses native Claude rather than API relay when Claude tab is open.
–Send button detection improved: aria-label cascade → data-testid → proximity heuristic.

v2.3.0

October 4, 2026

Minor

@all broadcast goes fully native across all 7 platforms.

Broadcasting

–@all now broadcasts to all 7 native agents simultaneously using Promise.allSettled().
–Synthesis prompt sent natively to Claude once all 7 responses arrive.
–Responses labelled by agent with coloured dot in the panel.

Agents

–Perplexity, Grok, Mistral, and Copilot added to native routing.
–Natural language relay: "tell ChatGPT to write me a poem" — no @mention required.
–Auto-chain: if any agent response begins with @claude: ... it is forwarded automatically. Depth limit: 1.

v2.2.0

October 4, 2026

Minor

Memory UI, Agent Roles tab, and Engram memory types.

Memory

–Memory tab in panel: shows stored memories grouped by type — preferences, facts, episodes.
–Filter input at top to search by keyword.
–Each memory card shows source agent → target agent + date.
–× on any card permanently deletes that memory from the backend.

Roles

–Agent Roles tab: assign Writer, Coder, Researcher, Analyst, Summarizer, or Translator to each agent.
–Roles are saved locally (chrome.storage.local) and persist across tabs and sessions.
–Role injected as system context before every outgoing message.

v2.1.0

October 3, 2026

Minor

Railway backend, 7-platform support, and Engram memory launch.

Platforms

–Expanded from 3 to 7 platforms: ChatGPT, Claude, Gemini, Perplexity, Grok, Mistral, Copilot.
–Background service worker (background.js) handles cross-tab routing via Chrome tabs API.
–@all broadcast introduced — sends to all available platforms simultaneously.

Backend

–Railway backend launched: /relay, /memory/list, /memory/store, /memory/delete.
–Engram memory layer: POST /memory/store after every exchange, GET /memory/search before every message.
–Free tier: 200 memory API calls/month. Native routing: unlimited, never counted.
–Cross-tab history: up to 60 messages stored in chrome.storage.local with 300ms debounce.

v1.0

October 3, 2026

Launch

First public release. ChatGPT, Claude, Gemini. No backend. @mention routing.

Launch

–Chrome extension with content script injected on ChatGPT, Claude, and Gemini.
–@mention routing: type @claude inside ChatGPT, Claude responds in a floating panel.
–No backend, no API key, no account required — fully local.
–Alt+Shift+C keyboard shortcut to open/close the panel.
–↗ button to paste full panel context into the host AI's input.

Always improving.

v2.4.0 is live. Follow development on GitHub, or add the extension and update automatically.