MCP Index Community Cleanup Queue

The cleanup queue is the fastest way to make the TensorBlock MCP Index more useful without adding a brand-new server.

Use it when you want a focused contribution: verify a stale entry, improve metadata, fix a category, close a duplicate, or help turn an automated health finding into a clean PR.

Open Task Lanes

Lane	Best for	Queue
Good first metadata	Small install, docs, auth, transport, license, or tool metadata fixes	Open good-first-metadata issues
Broken entries	Dead links, duplicate entries, stale projects, bad categories, or unsafe entries	Open broken-entry issues
Catalog health	Automated reports from the scheduled catalog health checker	Open catalog-health issues
Needs triage	New community reports that need routing or verification	Open needs-triage issues
Server submissions	New MCP server requests that need duplicate checks and category review	Open server-submission issues
Needs metadata	Server submissions missing required fields or category routing	Open needs-metadata submissions
Ready server PRs	Server submissions with automation-generated PRs ready for maintainer review	Open ready-for-pr submissions
Duplicate submissions	Server submissions matching an existing project URL	Open duplicate submissions
Client config requests	New MCP client or install target formats	Open client-config issues
Profile claims	Community maintainer claims for indexed profiles	Open claim-profile issues

How To Pick A Task

Start with a task that has a clear source link. Good cleanup issues usually include a TensorBlock profile URL, server id, project URL, or generated catalog-health marker.

Before editing, check:

• whether the project URL already appears elsewhere in docs/*.md,
• whether the public project README or docs confirm the proposed change,
• whether the fix belongs in a category markdown file or data/server-metadata/*.json,
• whether the issue already has an automation-generated cleanup PR or draft investigation PR.

For server submissions, use the intake labels to pick the next action:

• needs-metadata: ask for the missing URL, category, or description before editing docs.
• duplicate: close or link to the existing entry unless the submitter shows it is a distinct MCP server.
• ready-for-pr: review the generated PR instead of hand-editing the issue.
• automation-blocked: open the PR from the generated branch or fix workflow permissions, then rerun by editing the issue.

Maintainers can backfill these labels with the MCP Add Server Intake Refresh workflow. Use dry_run=true to inspect the planned changes across open server-submission issues, then rerun with dry_run=false to apply the labels. Set issue_number when you only want to refresh one issue.

Comment on the issue before doing larger cleanup work so two contributors do not handle the same task.

What To Change

Most cleanup PRs fall into one of these shapes:

• Edit a category page under docs/*.md to remove a duplicate, remove a verified dead link, fix a broken URL, or move an entry to a better category.
• Add or update a data/server-metadata/{serverId}.json sidecar for install, auth, transport, docs, license, client, tool, maintainer, or verification metadata.
• Update a generated investigation spec under docs/broken-entry-reports/ with the maintainer decision.
• Add client config support after a request spec under docs/client-config-requests/ is reviewed.

If you are fixing metadata, prefer structured labels in the issue or PR body:

Install: npx -y example-mcp
Transport: stdio
Auth: api-key, requires EXAMPLE_API_KEY
Docs URL: https://docs.example.com/mcp
License: MIT
Tools: search, fetch_profile
Tool count: 2

Validation

For docs-only cleanup, run the smallest useful check:

npm run catalog:build

For metadata, generator, API, or workflow changes, run:

npm run catalog:build
npm run profiles:build
npm test
npm run typecheck
npm run build

Generated files such as data/catalog.json, data/catalog-errors.json, and data/profiles/*.json are build outputs. Do not include them in a cleanup PR unless a maintainer explicitly asks for generated data.

Community Flow

The queue is meant to turn drive-by reports into visible contribution opportunities:

• issue forms add route labels and next-step comments,
• clear dead-link reports can create direct cleanup PRs, while ambiguous reports can create draft investigation PRs,
• catalog-health issues surface stale or broken entries proactively,
• merged server PRs post profile, API, install-config, and badge links,
• cleanup contributors can point maintainers to verified sources instead of leaving loose notes.

Join the TensorBlock Discord when a cleanup decision needs maintainer context or broader community discussion.