Bugra Karsli LogoBuğra Karslı
Back to Blog
May 23, 2026•3 min read

The Agent Manager Era: v3.2–v3.6

XerefAIAgent ManagerIntegrationsRelease Notes

Five releases that transform Xeref from a chat dashboard into a full agent management platform — with a live Agent Manager, Brand Audit, Anthropic webhook integration, and 6 new frontier models.

Between May 21 and May 23, Xeref shipped five releases that, taken together, represent the biggest surface area expansion since v2.0. The theme: agents you can actually manage, monitor, and build on.

v3.2 — The Agent Manager (Ctrl+E)

Pressing Ctrl+E now opens a full Agent Manager dashboard instead of a settings overlay. It ships with:

  • Live conversation history in a sidebar with unread indicators and status badges.
  • Brand Audit panel scoring xeref.ai against the CLAWS methodology (Connect, Listen, Archive, Wire, Sense, Identity) with a live alignment gauge, gap detection, and one-click copy remediation.
  • Provide Feedback modal — bug reports, feature requests, and general feedback with screenshot and log attachment support.
  • Knowledge Base view — a split-pane artifact browser with Copy Content, Copy Path, and Export Artifact in a ... context menu.
  • Workspace detail cards — agent count, task completion stats, active pipeline info, and an Open Editor button per workspace.
  • IDE integrations — Xeref is now natively connected to Antigravity IDE, VS Code (via GitHub Copilot Chat MCP extension), and Claude Code CLI via the self-hosted /api/mcp endpoint.

v3.3 — Three-Brain Stack and Screenshot Uploads

The Three-Brain Stack became official in this release: Antigravity IDE for MCP-powered day-to-day work, Claude Code CLI for deep cross-file refactoring, and Xeref.ai as the deployment and agent management surface. Screenshot upload in the feedback modal also went fully functional — inline preview, memory-safe cleanup on close.

v3.4 — Everything Goes Live

  • Live conversations: Real chat history from Supabase. Clicking a conversation fetches and displays the full message thread on demand.
  • Live artifacts: The Knowledge panel is backed by real artifact data.
  • Live workspace stats: Total Tasks, Completed, Pending, Active Pipelines, and Success Rate from real Supabase tables — not hardcoded.
  • Feedback persistence: Bug reports and feature requests are now stored in the database.
  • Anthropic Webhook receiver: A new endpoint at /api/webhooks/anthropic is ready to receive session.status_* events.

v3.5 — Agent Sessions and Webhook Hardening

The Sessions panel in Agent Manager lists all Anthropic Managed Agent sessions with live status dots and pass/fail badges. Clicking a session shows a full detail panel with timeline, an expandable Success Rubric, and the raw outcome payload.

The Anthropic webhook handler now verifies HMAC-SHA256 signatures with a 5-minute replay-attack window. Sessions persist to a new agent_sessions table with RLS restricting reads to the owning user.

v3.6 — Six New Models and MCP Setup Guides

Six models were added to the chat selector:

  • Pro: GPT 5.5 (OpenAI), Kimi K2.6 (Moonshot AI), GLM 5.1 (Z.AI), Grok 4.20 (xAI).
  • Ultra: GPT 5.5 Pro (deep reasoning, 5–10 min think time), Gemini 3.1 Preview.

The login page and Settings MCP Token card now include a two-step mcp-remote install guide compatible with Claude Desktop, Cursor, Windsurf, and all stdio MCP clients — with your token pre-filled and a one-click Copy config button.

These five releases move Xeref firmly into the category of agent infrastructure — not just a chat UI, but a platform for building, monitoring, and managing autonomous systems.