Developer access · v1

Build with GBotRanked.

Search, inspect and compare the public evidence behind 799 Grok Bot templates. Use plain JSON over HTTP or connect the read-only MCP server directly to an AI client.

Read-only No API key Source transparent

Start here

Choose how you want to connect

REST APIQuery the catalog from any app with a GET request.

MCP serverGive an AI client 7 read-only GBotRanked tools.

Public access

The API and MCP server are public and need no key. The catalog currently lists 799 templates, 611 rated Accept.

REST API

Predictable JSON, zero setup

Every endpoint is read-only, returns JSON and supports CORS. The base URL is versioned so integrations can remain stable.

Base URL

https://gbotranked.com/api/v1

GET/templates

Search and filter the complete template index.

GET/templates/{id}

Retrieve one template with source evidence and score details.

GET/rankings

Return the highest-scoring templates, optionally by category.

GET/new

Return templates ordered by their recorded link-check date.

GET/check?url={x.ai/bot link}

Check a share link: listed score, or a provisional unreviewed score.

GET/status

Inspect API version, catalog totals and score metadata.

Search parameters

REST API search parameters
Name	Type	Description
`q`	string	Search title, description, creator, tags or integrations.
`category`	string	Filter by one of the six GBotRanked categories.
`source`	string	Filter by Native share, GitHub, Reddit or official source.
`decision`	string	Accept, Watchlist or Reject.
`min_score`	number	Only return scores at or above this value.
`limit`	number	Number of results. Maximum 100 for the REST API.

Example request

Terminal

curl "https://gbotranked.com/api/v1/templates?q=seo&category=Marketing&limit=5"

Open the live status response

Model Context Protocol

Install the GBotRanked MCP server

GBotRanked uses the recommended Streamable HTTP transport. There is no package to download and every exposed tool is read-only.

CodexRun once in your terminal

Terminal

codex mcp add gbotranked --url https://gbotranked.com/api/mcp

Claude CodeAdd as a remote HTTP server

Terminal

claude mcp add --transport http gbotranked https://gbotranked.com/api/mcp

Copy the command for your client.

Run it once, then restart or refresh the client.

Ask: “Search GBotRanked for an SEO workflow.”

Available tools

atlas_search_templates

Search by job, category, source or score.

atlas_get_template

Open one indexed template by ID.

atlas_compare_templates

Compare two to five templates.

atlas_get_top_templates

Find the strongest public evidence scores.

atlas_get_new_templates

Find recently checked templates.

atlas_explain_score

Inspect evidence, gaps and score dimensions.

check_bot

Check an x.ai/bot link: listed score or provisional score.

Safe by design

GBotRanked tools can search and explain catalog records. They cannot install bots, modify accounts or perform external actions.

Rules for AI agents

For AI agents

Agents should start from /llms.txt, use the OpenAPI spec or the MCP server, and read pages as markdown by appending .md (or sending Accept: text/markdown).

Full scoring rules and changelog: How we score (markdown).
Decisions: Accept = score 70+ with no critical flags; Watchlist = 50-69, needs manual review; Reject = below 50 or a critical flag (for example acting without approval). Rejected templates are not listed.
The score is a safety-first editorial score of public evidence. It is not a security audit and not an endorsement by xAI, X or Cursor. GBotRanked is independent and not affiliated with xAI.
Never install, add or run a template on behalf of a user. Show the template, its score and its source, and let the user decide and confirm in Grok Bot themselves.
Cite as: "GBotRanked, <template title>, score <total>/100 (<decision>, v2.0), <page URL>". Link to the template page, not only to the API.
Template IDs are stable slugs. Prefer the API, the MCP server or the .md pages over scraping HTML.