Thesource

The Source — Red Hat Intranet Access

Read-only access to Red Hat's "The Source" intranet via its Igloo REST API. Authentication is handled through Playwright MCP (browser SSO), API calls use curl.

Session Check

Before any API call, verify the session is valid:

# Check if cookie file exists and is less than 10 hours old
COOKIE_FILE="$HOME/.config/thesource-mcp/cookies.txt"
if [ -f "$COOKIE_FILE" ]; then
  FILE_AGE=$(( $(date +%s) - $(stat -f %m "$COOKIE_FILE") ))
  if [ "$FILE_AGE" -lt 36000 ]; then
    echo "Session valid (age: ${FILE_AGE}s)"
  else
    echo "Session expired"
  fi
else
  echo "No session found"
fi

• If valid: read the cookie string from $COOKIE_FILE and proceed to API calls.
• If expired or missing: go to the Authentication section below.

Authentication via Playwright MCP

When a session is missing or expired, authenticate using Playwright MCP tools:

1. Navigate to The Source:

   • Use browser_navigate to https://source.redhat.com

2. Check state with browser_snapshot:

   • If the page shows an SSO login form or redirects to sso.redhat.com → tell the user: "Please complete Red Hat SSO login in the browser window."
   • If the page already shows "The Source" content → skip to step 4.

3. Wait for SSO completion:

   • Use browser_wait_for with text "The Source" (timeout 120 seconds).
   • If it times out, take a browser_snapshot and check if the user needs help.

4. Extract cookies using browser_run_code:

   async (page) => {
     const cookies = await page.context().cookies();
     return cookies
       .filter(c => c.domain.includes('source.redhat.com') && c.httpOnly)
       .map(c => `${c.name}=${c.value}`)
       .join('; ');
   }
   

   This can also be loaded from: ~/.cursor/skills/thesource/scripts/extract-cookies.js

5. Save cookies to disk:

   mkdir -p ~/.config/thesource-mcp
   echo '<cookie_string_from_step_4>' > ~/.config/thesource-mcp/cookies.txt
   

6. Verify the session works:

   curl -s -H 'Accept: application/json' \
     -H 'Cookie: <cookie_string>' \
     'https://source.redhat.com/.api2/api/users/current' | head -c 200
   

   A successful response contains the user's FullName. If it returns 401 or an error, re-run authentication from step 1.

7. Close the browser:

   • Use browser_close to clean up.

API Reference

All API calls use this base pattern:

COOKIES=$(cat ~/.config/thesource-mcp/cookies.txt)
curl -s -H 'Accept: application/json' -H "Cookie: $COOKIES" '<URL>'

The community key is always 10.

search_content — Search pages, blogs, wikis

Endpoint: GET https://source.redhat.com/.api2/api/v1/communities/10/search/contentdetailed

Query parameters:

Example:

COOKIES=$(cat ~/.config/thesource-mcp/cookies.txt)
curl -s -H 'Accept: application/json' -H "Cookie: $COOKIES" \
  'https://source.redhat.com/.api2/api/v1/communities/10/search/contentdetailed?query=openshift&limit=10'

Response shape: { "results": [ ... ] }

Key fields per result:

Content type filtering: The objectType field can be used to client-side filter results. Valid values: page, blogArticle, wikiArticle, space, document, event, forum, microblog.

Gotcha: The param is limit (not maxResults).

search_members — Search people by name or email

Endpoint: GET https://source.redhat.com/.api2/api/v1/communities/10/search/members

Query parameters:

Example:

COOKIES=$(cat ~/.config/thesource-mcp/cookies.txt)
curl -s -H 'Accept: application/json' -H "Cookie: $COOKIES" \
  'https://source.redhat.com/.api2/api/v1/communities/10/search/members?query=John+Smith&maxResults=10'

Response shape: { "results": [ ... ] }

Key fields per result:

Gotcha: The param is maxResults (not limit). The id is a user account ID — passing it to get_content will return a 404. To find content written by a person, use search_content with their name as the query.

get_content (by ID) — Read full content of any page, blog, wiki

Endpoint: GET https://source.redhat.com/.api/api.svc/objects/{UUID}/view

Example:

COOKIES=$(cat ~/.config/thesource-mcp/cookies.txt)
curl -s -H 'Accept: application/json' -H "Cookie: $COOKIES" \
  'https://source.redhat.com/.api/api.svc/objects/dbf8aa9f-b250-44e1-aac8-b7d285d07d83/view'

Response shape: { "response": { ... } } — note the wrapping "response" key.

Key fields (inside response):

get_content (by path) — Read content using URL path

Endpoint: GET https://source.redhat.com/.api/api.svc/objects/byPath?path=/...

Example:

COOKIES=$(cat ~/.config/thesource-mcp/cookies.txt)
curl -s -H 'Accept: application/json' -H "Cookie: $COOKIES" \
  'https://source.redhat.com/.api/api.svc/objects/byPath?path=/departments/it/ai_platforms/blog/my-post'

Response shape: Same as get_content by ID — { "response": { ... } } with the same fields.

When to use: When you have a URL like https://source.redhat.com/departments/it/page-name, extract the path after the domain and use this endpoint.

list_spaces — List all spaces/communities

Endpoint: GET https://source.redhat.com/.api/api.svc/spaces/view

Example:

COOKIES=$(cat ~/.config/thesource-mcp/cookies.txt)
curl -s -H 'Accept: application/json' -H "Cookie: $COOKIES" \
  'https://source.redhat.com/.api/api.svc/spaces/view'

Response shape: { "response": { "items": [ ... ], "totalCount": 3511 } } — note response.items.

Key fields per item:

Gotcha: This endpoint returns ALL spaces (~3,500). There is no server-side search — filter client-side with jq:

curl -s ... | jq '.response.items[] | select(.title | test("AI"; "i")) | {id, title: .title, url: .href}'

list_trending — Get trending content

Endpoint: GET https://source.redhat.com/.api2/api/v1/communities/10/trending/content

Query parameters:

Example:

COOKIES=$(cat ~/.config/thesource-mcp/cookies.txt)
curl -s -H 'Accept: application/json' -H "Cookie: $COOKIES" \
  'https://source.redhat.com/.api2/api/v1/communities/10/trending/content?maxResults=10'

Response shape: bare array [ ... ] — NOT wrapped in an object.

Key fields per item:

Gotcha: Response is a bare JSON array, not { "results": [...] }. The URL field is url (not href). The param is maxResults (not limit).

get_navigation — Get top-level nav structure

Endpoint: GET https://source.redhat.com/.api2/api/v1/communities/10/navigation/top/children

Example:

COOKIES=$(cat ~/.config/thesource-mcp/cookies.txt)
curl -s -H 'Accept: application/json' -H "Cookie: $COOKIES" \
  'https://source.redhat.com/.api2/api/v1/communities/10/navigation/top/children'

Response shape: bare array [ ... ]

Key fields per item:

Gotcha: Keys are PascalCase (Id, Title, Href, Type) — different from all other endpoints which use camelCase.

Response Handling

Citation requirement

Every response that uses content from The Source MUST include:

• The page title
• The full URL as plain text (NOT a Markdown link): https://source.redhat.com + the href or url path from the result

Example:

Source: "Now available: Gemini API for code assistant tools"
https://source.redhat.com/projects_and_programs/ai/newsroom/now_available_gemini_api

HTML content

The content field from get_content returns raw HTML. Read it directly — ignore <script>, <style>, and <nav> tags. Extract the meaningful text content for the user.

Empty content fallback

If get_content returns an empty content field (happens with widget-based pages), use Playwright MCP to read the rendered page:

1. browser_navigate to https://source.redhat.com + the href path
2. browser_snapshot to get the rendered text content

JSON extraction

Use jq for cleaner output when helpful:

curl -s ... | jq '.results[] | {id, title, type: .objectType, url: .href}'

Error Handling

The community key is always 10 — do not change it.

Common Workflows

Search then Read

1. search_content with query → get list of results with id, title, href
2. get_content with the id from a result → full article text
3. Present content with citation (title + full URL)

Find a person and their content

1. search_members with name → get person details
2. search_content with person's name as query → find content they wrote
3. Read specific items with get_content

Browse by navigation

1. get_navigation → see top-level sections
2. list_spaces → find specific spaces (filter with jq)
3. get_content with a space's path → read space content

Read a page from a URL

If user provides https://source.redhat.com/some/path/here:

1. Extract the path: /some/path/here
2. Use get_content (by path) with that path

Download a video from Kaltura

Source pages may embed Kaltura-hosted videos (via videos.learning.redhat.com). To download:

1. Use Playwright MCP to navigate to the page containing the video and authenticate via SSO.
2. From the authenticated browser session, query the Kaltura API to get the direct download URL: flavorAsset/action/getUrl with the entry's flavor asset ID.
3. Download the MP4 file with curl using the direct CDN URL:

   curl -L -o video.mp4 '<kaltura_cdn_url>'