Glama Status

Daily-refreshed Glama score tracker for the sandraschi fleet. Scrapes per-tool TDQS grades from glama.ai, stores in SQLite, surfaces via MCP tools and a web dashboard. Tracks deltas between snapshots, generates daily reports, and flags stale repos needing rescan.

Quickstart

just install
just web          # HTTP backend on :11072
just web-frontend # Vite dev on :11073
just web-dev      # Full stack
just refresh      # Manual scrape + snapshot

MCP Tools

Tool	Purpose
glama_status	Portmanteau: list, get, worst_tools, refresh, history, staleness, report, deltas
glama_scores_summary	Compact grade distribution and per-repo stats
glama_daily_report	Full markdown report with deltas, worst tools, stale repos

Web Dashboard

Route	Description
/	Sortable fleet score table
/api/repos	All repos with per-tool breakdowns
/api/repos/{name}	Single repo tool breakdown
/api/report	Full daily report JSON
/api/deltas	Score changes since last snapshot
/api/refresh (POST)	Trigger rescrape + snapshot
/api/worst-tools	Lowest-scoring tools fleet-wide

How it works

1. Scraper fetches glama.ai/mcp/servers/{author}/{repo}/score pages
2. Parser extracts per-tool grades, 6 TDQS dimension scores, coherence, and maintenance data
3. SQLite stores current scores + snapshot history for delta tracking
4. Daily refresh via Scheduled Task or just refresh creates a new snapshot
5. Delta engine compares latest two snapshots to show changes

Scored repos (10 of 35 registered on Glama)

Repo	Grade	Score	Tools
blender-mcp	C	2.70	67
windows-operations-mcp	B	3.00	17
virtualization-mcp	B	3.06	9
worldlabs-mcp	B	3.38	20
robotics-mcp	A	3.58	8
bumi-mcp	A	3.64	2
xkcd-mcp	A	3.67	6
cursor-mcp	A	3.80	6
steam-mcp	A	3.81	14
email-mcp	A	3.82	10

25 more registered on Glama but not yet analyzed (page exists, no tools).

Scoring (Glama TDQS)

Dimension	Weight	Description
Purpose Clarity	25%	First sentence states what the tool does
Usage Guidelines	20%	When to / not to call, preconditions
Behavioral Transparency	20%	Returns, side effects, error conditions
Parameter Semantics	15%	Every param: type, values, what it affects
Conciseness & Structure	10%	Not a wall of text, not a one-liner
Contextual Completeness	10%	Enough context to use without reading source

Server-level = 60% mean + 40% minimum - one bad tool pulls the whole score down.

Ports

• Backend (FastAPI + MCP HTTP): 11072
• Frontend (Vite dev): 11073