Read-It-Later Digester

You bookmark dozens of articles every week but never get around to reading most of them. Moltbot solves this by reading them for you — instantly summarizing content, extracting key points, and storing everything in searchable memory.

Just send a URL to Telegram. No commands, no tags, no extra apps. Moltbot handles the rest.

Prerequisites

• Moltbot running and connected to Telegram
• Browser or Fetch MCP tool — Moltbot needs the ability to open URLs and read page content
• Memory system enabled — Summaries are stored in memory for later retrieval. See Memory System

How It Works

The workflow is intentionally frictionless:

1. You come across an interesting article on your phone or desktop.
2. You forward or paste the URL into your Moltbot Telegram chat. No additional text needed.
3. Moltbot detects that the message is a bare URL.
4. It fetches the full page content using the browser/fetch tool.
5. It generates a summary (under 300 words) and extracts 3-5 key takeaways.
6. It saves the summary, takeaways, and original URL to its memory system under a "reading-notes" category.
7. It sends you the summary in the chat so you get immediate value.

Days or weeks later, you can ask natural language questions about anything you have saved:

You: What articles about RAG did I read last week?
You: Summarize everything I've saved about Rust async
You: Find that article about database indexing strategies

Vector search means you do not need to remember exact titles or keywords. Semantic similarity handles the matching.

Setup

Step 1: Add the URL Handling Rule to SOUL.md

Add the following to your SOUL.md file (Moltbot's behavioral configuration):

## Link Handling

When the user sends a URL with no additional text:
1. Use the browser tool to open and read the full page
2. Generate a summary under 300 words
3. List 3-5 key takeaways
4. Save the summary and URL to memory under "reading-notes"

This rule tells Moltbot how to behave when it detects a bare URL. The AI interprets the rule and applies it contextually — you do not need to handle URL parsing yourself.

Step 2: Test with a URL

Send any article URL to your Moltbot chat:

https://example.com/blog/some-interesting-article

You should receive a summary back within a few seconds. If the page is behind a paywall or JavaScript-heavy, results may vary (see Edge Cases below).

Step 3: Customize the Summary Format

You can adjust the SOUL.md rule to match your preferences:

Shorter summaries with tags:

## Link Handling

When the user sends a URL with no additional text:
1. Fetch and read the full page
2. Generate a 2-3 sentence summary
3. List 3 key takeaways as bullet points
4. Assign 2-3 topic tags (e.g., #ai, #rust, #infrastructure)
5. Save to memory under "reading-notes" with the tags

Academic paper handling:

## Link Handling — Academic Papers

When the user sends an arXiv or academic paper URL:
1. Fetch the page (or PDF abstract)
2. Summarize the paper's contribution in 3-4 sentences
3. Note the methodology and key results
4. Rate the practical relevance: high / medium / low
5. Save to memory under "papers"

With critical analysis:

## Link Handling

When the user sends a URL with no additional text:
1. Fetch and read the full page
2. Summarize in under 300 words
3. List 3-5 key takeaways
4. Add a "Critical Notes" section: any claims that seem unsupported, biases, or missing context
5. Save to memory under "reading-notes"

Step 4: Set Up a Weekly Reading Digest (Optional)

Combine with a cron job to get a weekly roundup of everything you saved:

cron:
  - name: weekly-reading-digest
    schedule: "0 10 * * 0"
    channel: telegram
    prompt: |
      Search my reading-notes in memory from the past 7 days.
      Generate a digest:
      - Group articles by topic
      - For each article, include the title, URL, and a one-line summary
      - End with "Top 3 Most Important" picks based on your judgment
      Keep the total under 500 words.

Edge Cases and Troubleshooting

• Paywalled content: If an article is behind a paywall, the browser tool may only retrieve a snippet or the paywall notice itself. Moltbot will do its best with the available text, but the summary will be limited. Consider sending the article text directly if you have access.
• JavaScript-heavy single-page apps: Some sites render content entirely via JavaScript. Basic fetch tools may return empty or boilerplate HTML. A headful browser MCP tool handles these better.
• PDF links: If the URL points to a PDF, behavior depends on your MCP tool's capabilities. Some browser tools can extract PDF text; others cannot. Test with a PDF link to verify.
• Very long articles (10,000+ words): The summary will still be constrained to your configured length, but the LLM may hit context window limits when processing very long pages. The browser tool typically truncates at a reasonable length.
• URLs sent with commentary: The SOUL.md rule specifies "URL with no additional text." If you send a URL with a comment like "Check this out," Moltbot may treat it as a normal conversation rather than triggering the read-later flow. To force it, send the URL alone on its own message.
• Duplicate URLs: If you send the same URL twice, Moltbot will process it again and create a second memory entry. You can add a rule: "Before saving, check if this URL already exists in memory. If so, skip saving and just show the existing summary."

Pro Tips

• Ask follow-up questions. After receiving a summary, you can ask deeper questions: "What specific benchmarks did they use?" or "How does this compare to the approach in that other article I read about vector databases?" Moltbot will use both the fetched content and its memory to answer.
• Build a personal knowledge base. Over weeks and months, your reading-notes memory becomes a searchable knowledge base. Use it for research: "What have I read about LLM fine-tuning techniques in the past 3 months?"
• Tag articles manually for better retrieval. After the summary appears, add your own note: "Tag this as important for the Q3 architecture review." This enriches the memory entry and makes targeted searches easier.
• Combine with the Writing Assistant. When drafting an article or report, ask Moltbot to pull relevant reading notes: "Find all my saved articles about self-hosting and summarize the key arguments." See Writing Assistant.
• Share digests with your team. If you use Moltbot in a group chat, the read-later flow works the same way. Team members can all contribute URLs, and the weekly digest becomes a shared reading list.

Related Pages

• Memory System — How saved articles are stored and searched
• MCP Tools — Setting up browser and fetch tools
• Personalized Daily Briefing — Include reading highlights in your morning summary
• Writing Assistant — Use saved articles as research material
• Scheduled Tasks — Setting up the weekly digest cron