Skip to main content

Release Notes — v1.3.0

June 30, 2026

This is a major release. It introduces Agents — configurable, first-class AI assistants you can tailor and chat with — alongside guided onboarding for new users, organization- and instance-level system prompts, a faster and more capable report pipeline, and spreadsheet tooling.

New Features​

Agents​

Agents are now a first-class feature. An Agent is a configured AI assistant with its own permissions, context, and behavior, packaged so you can chat with it directly instead of assembling the same setup by hand each time.

This release ships Agents end to end:

  • Configure an Agent — define its permissions, schema, and context.
  • Manage your Agents — a dedicated management UI to create, edit, and organize them.
  • Chat with an Agent — the Agent's configuration is applied automatically during chat execution, so a conversation starts already set up the way you defined it.
  • Recommended prompts — when you open an Agent, suggested starter prompts surface so you know what it's good at and can get going immediately.

Agents are rolling out gradually — availability is gated per organization, and the Create Agent action requires organization management permissions. Contact your Revelry Ai team if you'd like Agents enabled. Full Agents documentation will follow as the feature becomes broadly available.

Guided Onboarding​

New users are now welcomed with a guided onboarding chat interface and a redesigned landing experience. An account-setup flow walks first-time users through getting started, replacing the previous cold-start into an empty workspace.

Organization & Instance System Prompts​

You can now configure a system prompt at the organization level, applied across that organization's chats, and administrators can set an instance-level system prompt that applies across the whole instance. This lets teams and operators establish consistent default behavior, tone, and guardrails without editing every Skill or chat.

Spreadsheet Tooling​

Chat can now work with spreadsheets. A new spreadsheet tool — including a spreadsheet_analyze capability available to subagents — lets the AI read, analyze, and reason over spreadsheet data as part of a conversation.

Improved Reports​

Report artifacts got a substantial upgrade this release:

  • Higher-fidelity PDFs — report PDFs are now rendered through headless Chromium in a sandboxed environment, producing output that more faithfully matches the on-screen report.
  • Chat images in reports — images you upload in a chat now render directly in generated reports.
  • Copy & edit the HTML source — you can view and edit the raw HTML behind a report artifact for fine-grained control over the output.

Artifact Sharing via MCP​

New MCP tools let the AI share artifacts directly, so sharing can happen as part of a conversation or workflow rather than as a separate manual step.

Data-Source-Aware Knowledge Tools​

The Data Center (Knowledge) MCP tools are now aware of the data sources available to you, so the AI can make better decisions about where to look when answering questions from your Knowledge.

Credit Balance in the Nav​

Your organization's credit balance now appears in the top navigation, so usage is always visible at a glance.

Super Admin — Credit Plans​

Super admins get a new UI to create, archive, and delete Credit Plans — both regular and freemium — with active subscription counts and audit logging.

UI Improvements​

  • LED theme — a new LED theme with refreshed pills and favicons.
  • Onboarding polish — visual refinements throughout the new onboarding flow.
  • Categories & sidenav — categories are decoupled from artifact types, with Skill support added and an updated side navigation.
  • Smaller refinements: manual textarea resizing is preserved, Agent featured-prompt textareas auto-resize, overflowing pill labels truncate cleanly, the thinking block has more breathing room before the response, and a non-functional placeholder voice-record button was removed.

Security​

  • Expanded Sentry scrubbing — additional scrubbing of sensitive data from error reports, including a new Markdown scrubber.
  • No password re-use — you can no longer set your current password as the "new" password during a change or reset.

Performance & Reliability​

  • Full project overview reaches the assistant — the silent 2,000-character cap on injected project overview context was removed; the entire overview is now available to the AI.
  • PDF downloads are more robust — the report PDF download button now waits up to 5 minutes to cover a full sandbox render, and a guard ensures a render crash notifies you instead of leaving the download hanging.

Bug Fixes​

  • skill_execute now validates its input parameters and rejects invalid ones.
  • The list_categories MCP tool now exposes subcategories.
  • Artifact URLs are retained in subagent replies regardless of the popout flag.
  • A project's Skills tab is now correctly scoped to that project's categories.
  • Fixed a transparent overlay in the LED theme so the artifact panel renders a solid background at all sizes.
  • Onboarding data entered before registration is no longer dropped on the way to the registration form.
  • Resolved layout issues including orphaned dividers and edit-in-place title overlap.

Under the Hood​

  • Migrated the Markdown renderer from earmark to mdex.
  • Added a centralized ResourceScopes module for permission handling.
  • Upgraded Elixir, OTP, Debian, WeasyPrint, and Node (25 → 26), plus routine dependency and CI maintenance.
  • Extensive feature-flag gating to scope premium features appropriately across tiers, including the in-chat skill selector.