At a glance

InterfaceAddressAuth
MCP server (Streamable HTTP)https://www.masamedia.co.il/mcp/None
REST API, v1 (JSON)https://www.masamedia.co.il/v1/None
OpenAPI 3.1/openapi.jsonNone
Any page as MarkdownAdd .md to the URL, or send Accept: text/markdownNone

Masa Media Digital is an SEO and GEO agency in Israel. The content is mostly Hebrew; field names, error messages and this page are in English. Quote the content with a link to the page it came from.

MCP server

A public, read-only MCP server at https://www.masamedia.co.il/mcp/ (with the trailing slash). It is stateless, needs no authentication and has these tools:

  • search: Search the published pages of masamedia.co.il (services, Hebrew SEO and GEO articles, glossary, case studies). Returns up to 10 results with id, title and URL; pass an id to fetch to read the page.
  • fetch: Read one page of masamedia.co.il as Markdown. Pass the id (URL) returned by search, or any path on the site such as /seo-pricing/.
  • get_company_info: Facts about Masa Media from its own site: services (with page URLs), published prices (monthly SEO packages in ILS without VAT), contact details, or client case studies with their results.
  • glossary_lookup: Find terms in the Hebrew SEO glossary of masamedia.co.il. Pass part of a term (Hebrew or English) or leave it empty to list terms alphabetically.
  • google_updates: List Google's official named ranking updates from the last 12 months, newest first, with start and end dates and a link to Google's announcement.

search and fetch follow the format ChatGPT uses for deep research, so results come back with URLs it can cite.

Connect a client

  • Claude: Settings, Connectors, Add custom connector, and paste the server URL.
  • ChatGPT: in developer mode, create an app for a remote MCP server with no authentication.
  • Claude Code: claude mcp add --transport http masa-media https://www.masamedia.co.il/mcp/
  • MCP Inspector: npx @modelcontextprotocol/inspector, transport Streamable HTTP.

REST API

Base URL https://www.masamedia.co.il/v1/. GET only, JSON responses, described in /openapi.json.

RequestReturns
/v1/search/?q=GEOSearch published pages. Parameters: q (required), type, limit (1-20), cursor or offset.
/v1/page/?url=/seo-pricing/One page as Markdown, with its title and canonical URL.
/v1/services/The six services, with page URLs. Optional id for one service.
/v1/pricing/The monthly SEO packages as the homepage lists them (ILS, VAT not included).
/v1/contact/Phone, email, address, opening hours.
/v1/case-studies/Client case studies with their headline result. Paged.
/v1/glossary/The Hebrew SEO glossary. Paged, optional q filter.
/v1/google-updates/Google's official ranking updates from the last 12 months. Optional type: core, spam or other.

Example (response of 30 September 2026):

curl "https://www.masamedia.co.il/v1/search/?q=GEO&limit=1"

{
  "query": "GEO",
  "total": 40,
  "limit": 1,
  "offset": 0,
  "nextOffset": 1,
  "nextCursor": "bzox",
  "results": [
    {
      "id": "https://www.masamedia.co.il/ai-seo/geo/",
      "title": "קידום GEO: איך גורמים למותג להופיע ב-ChatGPT, בג'מיני ובמצב AI?",
      "url": "https://www.masamedia.co.il/ai-seo/geo/",
      "type": "service",
      "excerpt": "מה זה קידום GEO, ואיך מכניסים את המותג לתשובות של ChatGPT, ג'מיני, מצב AI וסקירות ה-AI של גוגל? מדריך מעשי לקידום במנועי תשובות.",
      "updatedAt": "2026-09-16T06:56:40.046Z"
    }
  ]
}

Pagination

Search, case studies and the glossary are paged. They take limit and return total and nextCursor; pass cursor=<nextCursor> to get the next page, until nextCursor is null. A plain offset works too.

Errors

Every error is an RFC 9457 problem document (application/problem+json) that says what to change:

{
  "type": "https://www.masamedia.co.il/developers/#errors",
  "title": "Invalid query",
  "status": 400,
  "detail": "Pass q with 2 to 120 characters, for example ?q=GEO"
}

Rate limits

Up to 120 requests per 60 seconds per client on /v1/ and /mcp/. Responses carry RateLimit-Policy and RateLimit headers; over the limit the answer is 429 with Retry-After.

Versioning

The major version is in the path. Changes inside v1 only add fields or endpoints. A breaking change will ship as /v2/, and v1 will be retired only after Deprecation and Sunset headers on v1 responses have announced the date.

Markdown for agents

Every page has a Markdown version with the page content only (no navigation, forms or widgets). There are two ways to get it:

  • Add .md to the address: /seo-pricing/ becomes /seo-pricing.md, and the homepage is /index.md.
  • Request the normal URL with Accept: text/markdown. Browsers and search engines keep getting HTML; the format is never chosen by user agent.

Markdown responses point to the HTML page with Link: rel="canonical". Section indexes live at /<section>/llms.txt, for example /blog/llms.txt.

WebMCP

In a browser that supports WebMCP, every page registers read-only tools for an agent working on it: search_site, get_pricing and get_contact_info. On the homepage, the contact page and the English page, the contact form is also a tool, request_consultation: the agent fills it in, and the visitor reviews it and presses send. Nothing is sent without that click.

Discovery files

Open-source tools

Masa Media also publishes open-source SEO and GEO tooling (an audit engine, a CLI, a browser extension and a WordPress plugin) on GitHub.

Questions

Write to contact@masamedia.co.il. To work with Masa Media, use the contact form or call +972-72-338-4444.