Skip to main content

Scorecard tiers, newest catalog first, and MCP tool hints

· 4 min read
Venue.sh Team
Venue.sh Team
Venue.sh

This release retires the old pass/fail scorecard rule. You now define ordered tiers, and an entity or repository earns the highest tier it can actually clear. The catalog table lists recently changed entities first, and tools on the Venue MCP Endpoint carry annotations so clients such as Cursor and Claude Code can tell a read-only lookup from something that writes.

Scorecard rules now use ordered tiers​

A scorecard rule is no longer a single criterion that every target either passes or fails. You define a stack of tiers. Each tier has its own description and criteria. For repository-content rules, each tier can also have its own file match. An entity or repository achieves a tier only if that tier and every lower one passed.

  • Tier achievement - The Scorecards list replaces the old Compliance Score chip and Pass / Fail column with a Tier achievement bar (how many targets reached each tier) and a Next target column (the first unmet tier and its description).
  • Entity page - Open a catalog entity's Scorecards tab to see that entity's own stepper and next-target progress, not the rule-wide totals.
  • Repository content - Build tiers with category templates, assisted generation from the tier's description, or by hand. Preview still evaluates every tier against the same fetched snapshot.
  • Missing files fail - If a repository-content tier's file match finds nothing, that tier fails. It does not skip the rest of the stack.

Open Scorecards to create or edit a multi-tier rule, then check the list and a catalog entity's Scorecards tab. See Scorecards and Expressions for how evaluation works.

MCP tools advertise how they behave​

Tools on the Venue MCP Endpoint now include annotation hints. Your MCP client can treat most Venue tools as read-only and idempotent. A few stay closed-domain: user info, template JSON Schema, and template validation.

These are hints for the client. They do not change what the tools do when you call them. The hints include:

  • read-only - The tool is read-only and does not modify data.
  • destructive - The tool may modify data (e.g. update or delete).
  • idempotent - The tool is idempotent and can be called multiple times without changing the result.
  • open-world - The tool may access data outside of the Venue.sh data model (e.g. external APIs or the internet).

See Venue MCP Endpoint for how to connect Cursor, Claude Desktop, or another MCP client.

Fixes and improvements​

  • Skills in entity pages - Skills now appear on entity pages, not just in the Skills tab.
  • MCP Apps - The Venue.sh MCP Server now exposes MCP Apps so that AI clients (e.g. Claude, Claude Desktop, VS Code GitHub Copilot) can visualize the entity data and click through to the entity page.
  • Consistent dates - Timestamps across the dashboard now use yyyy-MM-dd or yyyy-MM-dd HH:mm:ss. Relative phrases such as "3 minutes ago" are gone, including on agent, MCP server, skills, and plugin entity tabs.
  • Settings copy - AI Settings now describes general AI usage for your organization, rather than repository summaries alone. See Settings and AI Settings.

See Also​

  • Scorecards - Define catalog and repository rules, including tiers and preview.
  • Expressions - Operators and fields used in scorecard criteria.
  • Importing Repositories - Import repositories so they appear as catalog entities.
  • Venue MCP Endpoint - Connect AI tools to your Venue.sh data over MCP.
  • Settings - Account and organization configuration, including AI Settings.
  • AI Settings - Index repositories for search, general AI usage, and Skills discovery.