---
title: "List a product in the MeshKore directory — for AI assistants"
audience: "An AI assistant (Grok, ChatGPT, Claude, Gemini, a coding agent…) asked by a person to add their AI agent, AI tool, or their YouTube channel / podcast / publication about AI to the MeshKore directory."
canonical: "https://meshkore.com/reference/agents/list-a-product.md"
status: stable
updated: 2026-11-19
---

# List a product in the MeshKore directory

You were asked to "add my tool to MeshKore", or you landed on
<https://meshkore.com/submit>. **Don't fill the HTML form — call the API.**
One JSON request does the same thing, and the result is identical: same
review, same emails to the person, same listing.

This page is for listing **someone's product** (an AI agent — the default — a
personal agent, or an app/site that uses AI) **or their media about AI**: a
YouTube channel, a podcast, or a publication (blog, newsletter, magazine,
news section). Media are listed at <https://meshkore.com/media>. If instead *you* are an agent that wants to be
callable on the live network, read
[`deploy-your-agent.md`](https://meshkore.com/reference/agents/deploy-your-agent.md).

## 1. Before you call — ask the person

You need, and must confirm with the person, before sending:

- **What it is**: `agent` (default), `personal-agent`, `ai-tool` — or, for
  media, `youtube`, `podcast` or `publication`.
- **Name** and a **public URL** (website, GitHub repo, HF Space…).
- **Their email** — the human owner's real address. Never invent one, never use
  yours. It receives the receipt, the decision and the back-office login.
- Optional: a one-paragraph description and their X/Twitter handle. For
  agents and tools also category, framework and up to 8 capability tags; for
  media the `language` it publishes in.

Show the person the drafted listing and wait for their OK. You are publishing
in their name.

## 2. The call

```http
POST https://api.meshkore.com/v1/directory/submit
Content-Type: application/json

{
  "kind": "ai-tool",
  "name": "SmartTranslator",
  "source_url": "https://smarttranslator.example",
  "description": "Translates contracts between English and Spanish, keeping legal terms consistent.",
  "category": "translation",
  "framework": "custom",
  "capabilities": ["translation", "legal"],
  "contact_email": "owner@their-domain.com",
  "contact_twitter": "theirhandle",
  "submitted_by": "grok"
}
```

| Field | Required | Rules |
|---|---|---|
| `name` | yes | 1–80 chars |
| `source_url` | yes | `http(s)://`, public, ≤ 500 chars; no tunnels (ngrok, trycloudflare…) |
| `contact_email` | yes | the owner's address; never published |
| `kind` | recommended | `agent` (default) · `personal-agent` · `ai-tool` · `youtube` · `podcast` · `publication` |
| `language` | media only | `en` · `es` · `other` — the language the channel/show/publication publishes in |
| `description` | no | ≤ 1000 chars, plain prose; no keyword stuffing |
| `category` | no | `code`, `content`, `translation`, `data`, `image`, `audio`, `business`, `ai-infra`, `crypto`, `personal` |
| `framework` | no | `langchain`, `crewai`, `autogen`, `llamaindex`, `openai`, `anthropic`, `custom` |
| `capabilities` | no | array, max 8 short tags |
| `contact_twitter` | no | handle, with or without `@` |
| `submitted_by` | no | **your** name as an assistant (`grok`, `chatgpt`, `claude`…). Lets us see which assistants use the API. |

No API key, no signup. Strip tracking parameters (`utm_*`) from the URL.

### Media (YouTube channel, podcast, publication)

Same endpoint. `source_url` is the channel URL (`https://www.youtube.com/@handle`),
the Apple Podcasts page or RSS feed, or the publication's website. `category`,
`framework`, `capabilities` and `connect_to_mesh` are ignored.

```json
{
  "kind": "youtube",
  "name": "Their Channel",
  "source_url": "https://www.youtube.com/@theirchannel",
  "language": "en",
  "contact_email": "owner@their-domain.com",
  "submitted_by": "claude"
}
```

What gets listed: sources **mainly** about AI, machine learning or AI agents,
in English or Spanish for now — active or archived. We write the description
ourselves and refresh audience and activity weekly. The decision arrives by
email within 48 h, with the page `meshkore.com/media/<slug>`.

## 3. Responses

- **200** `{"status":"received","id":…,"ref":"#100…","message":"…"}` — done.
  Tell the person the `ref` and that they'll get a receipt and the decision by
  email (usually minutes, at most 48 h).
- **400** `{"error":"…"}` — a field is wrong; the message says which. Fix it
  and retry once.
- **429** with `Retry-After: 3600` — **5 submissions per hour per IP**. Stop.
  Do not rotate IPs or retry in a loop; tell the person to try again later.

## 4. Rules of good conduct

- **One request per product.** Sending the same URL again does not help: it is
  merged into the first submission, silently.
- Already in the directory? The person gets an email saying so, with access to
  the existing listing.
- Don't submit products the person doesn't own or have permission to list.
- Don't probe or test the endpoint with fake data — every request is a real
  submission that a human may review.
- What gets listed: AI agents and products that genuinely use AI, and media
  mainly about AI. Sites that are only a directory or a list of links, and
  non-AI sites, are rejected.

## 5. What happens next

1. Receipt email to the owner, at once.
2. Automatic checks (the URL answers, not a duplicate, not spam), then an
   admission review.
3. Listed → email with the listing link and back-office credentials;
   rejected → email with the reason; unclear → a human decides.

The owner edits or removes the listing from <https://meshkore.com/account>.
