Servers grouped by what they actually do. Every topic shows how many are in the registry and how many answer right now — the second number is the one that matters.
Servers that drive a real browser: open pages, click, fill forms, take screenshots. The heavy ones run Chromium somewhere, so whether the endpoint answers matters more here than anywhere else.
Search APIs wrapped for agents. Most need a key, so a 401 response here means working, not broken — we mark those separately.
Repository context, pull requests, code search and review. The largest group in the registry, and the one with the most abandoned projects.
Query and inspect databases from an assistant. Almost all require credentials, and the ones that do not are worth a second look before you connect them.
Vector stores and retrieval pipelines for grounding answers in your own documents.
Deployments, containers, cloud accounts. These act on live infrastructure, so check what tools a server exposes before granting it access.
Logs, traces, alerts and incidents pulled into the assistant while you debug.
Scanning, secrets and compliance checks. Worth reading the tool list closely — a security server usually asks for broad access by design.
Test runs, browser QA, coverage reports and visual checks.
Issues, sprints and roadmaps from Jira, Linear, Asana and friends. Mostly hosted servers with OAuth.
Workspace pages, wikis and personal notes as a source of context.
Messaging and mail. These can write on your behalf, so the tool list is the thing to read first.
Local filesystems and cloud drives. The local ones cannot be pinged — they have no network endpoint at all.
Spreadsheets, analytics and BI tools for asking questions of your numbers.
Market data, payments and accounting. This topic has the highest share of third-party wrappers using a vendor's name — check who actually runs the server.
Design files and UI context for handing work over to a coding agent.
Multi-step automations and app actions, usually fronting hundreds of integrations at once.
Parsing, extraction and OCR for documents an assistant cannot read on its own.
Publishing, listening and audience research across social platforms.
Stores, products and orders — mostly official vendor servers with OAuth.
Geocoding, places and forecasts. Many are open with no key required.
Crawling and structured extraction from pages, separate from full browser control.