Skip to main content
Glama
Browse API reference

Build with the MCP directory

Search Glama's curated catalog of open-source servers and remote connectors, inspect their capabilities, and keep your own directory synchronized.

Base URLhttps://glama.ai/api/mcp

Bearer authentication / 100 requests per second / cursor pagination

Quick start

Export your key, request the connector catalog, then pass each page's endCursor back as after until hasNextPage is false.

export GLAMA_API_KEY='glm_...'

curl --request GET \
  --url 'https://glama.ai/api/mcp/v1/connectors?first=10&sort=featured%3Adesc' \
  --header "Authorization: Bearer $GLAMA_API_KEY"

The MCP API provides access to our curated directory of MCP servers. Use it to integrate the Glama MCP server directory into your application.

Authentication

Directory read endpoints require an API key. Create one here and send it as Authorization: Bearer <key>.

The usage telemetry endpoint is unauthenticated so MCP clients can report tool usage without a Glama account.

Rate limits

Requests are limited to 100 per second per IP address, counted across every Glama API. Ten consecutive rejections earn a temporary ban.

Every response carries the IETF draft headers, so a client can pace itself without guessing:

Header

Meaning

RateLimit-Limit

Requests permitted per window.

RateLimit-Remaining

Requests left in the current window.

RateLimit-Reset

Seconds until the window resets.

Over the limit you get 429 with code: rate_limit_exceeded; wait RateLimit-Reset seconds and retry. Keep going and you get 403 with code: forbidden, which does not clear on its own — contact us if you hit it by mistake.

The limit is generous relative to what a bulk consumer needs: at the maximum page size of 100, mirroring the entire connector directory is a few hundred sequential requests.

Attribution is required

IMPORTANT

This data is licensed, not public domain. Under theAPI Data License, every page, screen, or view that displays data from this API must carry a visible credit to Glama, and every server or connector you present must link to its own listing on Glama.

1. Credit Glama on every page

  • Place the credit on each page that shows the data, not only on an "about" or "credits" page.

  • Give it prominence comparable to any other data-source or partner credit you display.

  • Link it to https://glama.ai without rel="nofollow", rel="sponsored", or rel="ugc".

Wording is up to you, so long as it names Glama. For example:

<a href="https://glama.ai">MCP data from Glama</a>

Every card, row, search result, or detail page devoted to a single server or connector must include a link to that record's listing on Glama — the url the API returns for it, not a copy of the listing you host.

<a href="https://glama.ai/mcp/servers/microsoft/playwright-mcp">Playwright MCP Server on Glama</a>

This is an additional link, not a replacement: whatever the card already links to — your own detail page, the repository, the connector endpoint — stays as it is, and the Glama link goes alongside it.

The same rules apply as to the credit: a standard crawlable link, reachable without a hover or a menu, and no rel="nofollow", rel="sponsored", or rel="ugc".

Building something where a visible credit does not fit, such as an internal tool, a research dataset, or a voice or CLI surface? Attribution can be waived under a commercial licence. Talk to us.

Data Sources

Our directory is built from multiple sources:

If you need a feature that isn't available, please contact us.

Endpoints

Connectors

Remote MCP servers: endpoints operated by their publisher that a client connects to over the network.

Where Servers describes source code you can run, a connector describes a running deployment: its URL, transport, authentication scheme and the tools it advertises.

Only approved connectors reachable over streamable HTTP are listed.

Instances

MCP server instances hosted by Glama.

These instances belong to users and are not publicly accessible.

Telemetry

Anonymous MCP tool usage reporting.