How to submit an MCP server to TensorBlock awesome-mcp-servers

You add a server here by writing one Markdown bullet in a category file. A build then turns that bullet into a profile page, an install-config preview for several clients, and a badge for your own README^[1]. That makes the line you write the thing the index says about you, so it is worth writing the line well.

A pull request or an issue form

• A pull request that adds the bullet to the right category file under docs/. The issue form itself says a pull request is still the fastest path^[2].
• The Add MCP server issue form, when you are unsure of the category or want help shaping the entry. It asks for a name, a project URL and a best category from a list of 13, a description of what an agent can do with it, and optionally install steps, transport, auth, clients and licence^[2]. When the form has enough, automation can open a draft pull request for you^[1].

The bullet, and where it goes

- [Example Weather](https://github.com/example-user/weather-mcp): Gets forecasts and weather alerts for a city. Install: `npx -y weather-mcp`.

Search the repository for your project URL first to avoid a duplicate, then add the bullet to the category page that fits^[1]. The README lists 33 categories with their counts. On 1 Oct 2026 the largest were AI & LLM Integration (1,632 entries), Knowledge Management & Memory (572), Finance & Crypto (445) and Developer Productivity & Utilities (414), and the whole index held 7,747 unique links^[1].

The metadata that makes the profile useful

A simple bullet is enough to be merged. The README says the best entries also answer: what can an agent do with this server, how is it installed or connected, whether it uses stdio, sse or streamable-http, whether it needs an API key, OAuth, a bearer token or nothing, which clients it supports, and where the docs, source, licence and public endpoint are^[1]. Extra structured values go in sidecar files the indexer reads, or through the metadata issue form. You do not need to touch those to be listed.

What you get after a merge

After a pull request lands on main, a deploy rebuilds the catalog and profiles. The README says a follow-up comment then gives you the public profile address, the API profile address, install-config links and the badge to paste into your README^[1]. Claiming the profile is a separate issue form, and the README is careful to say a claim is not the same as TensorBlock verification^[1].

Edits that get bounced

• Editing generated files. data/catalog.json and data/profiles/*.json are produced by the indexer. For a simple entry, editing the category page is enough^[1].
• A duplicate. Submissions through the form get an intake label, and duplicate means your project URL is already in the catalog. The others are needs-metadata, ready-for-pr and automation-blocked^[1].
• Missing fields. needs-metadata means a required field, a valid project URL, or category routing is still missing^[1].

By hand, the short way

Search, pick the category page, add the bullet at the end, and open a pull request. The title the form uses is "Add MCP server: " followed by the name^[2], and a pull request title in that style fits.

Current status, 1 Oct 2026

On 2 Oct 2026 the repository had 7 open pull requests and 2 open issues^[3][4], and 4 open issues labelled ready-for-pr on the day I read it^[5]. The index serves a free public API that other tools can read, described in the README^[1].