Inoreader logo Inoreader automation · runs daily

Triage unread Inoreader articles with an AI agent

You follow 80 feeds, and by Monday the AI News folder shows 400 unread articles. Most are the same story told five times, and the three you need are somewhere in the middle.

Cotlet does it in your own Inoreader account with get_unread_counts, list_articles, star_article and mark_as_read, called by your AI agent, an n8n workflow or a script.

Hosted · nothing to install or maintain

Can Inoreader do this on its own?

Inoreader offers rules and filters for this job (checked 17 Oct 2026). A rule can tag or mark as read by keywords in title and content, author, URL, language or RSS categories, and a duplicate filter removes copies by URL or title. Neither reads an article and judges whether it matters to you. The API can read and mark articles, but it needs a Pro plan, your own app and 100 requests per day per zone by default.

How you do it in Inoreader today

  1. Open the folder in the left sidebar and read the unread articles in the list, newest first.
  2. Select several articles with Shift + click, or switch on the Bulk actions toolbar button, which acts on up to 50 items at once.
  3. Mark the selection as read, tag it or save it to Read later from the same toolbar.
  4. Repeat for every batch of 50, because Inoreader has no way to say "star everything about MCP" in one step across a backlog.
  5. The backlog grows again with every feed update, so the same sorting is due again the next morning.

How to triage unread Inoreader articles with an AI agent

Prompt

Go through my unread Inoreader articles from the last 24 hours in the folder AI News. Star the ones about agents or MCP, mark the duplicates and obvious fluff as read, and give me a list of the starred ones with one line each.
  1. 01

    Size the backlog

    The agent calls get_unread_counts first, so it knows how many unread articles sit in the folder and does not page through a feed that has 3.

  2. 02

    Read only the new part

    list_articles runs with unread_only and a since time of 24 hours, newest first, and the agent follows the cursor until the window is covered.

  3. 03

    Judge each article

    The agent compares title and summary with the topics in your prompt. It groups articles that tell the same story and keeps the one from the source you named as preferred.

  4. 04

    Star the keepers

    star_article runs for the articles that match, so they stay in your Starred list even if you stop the run halfway.

  5. 05

    Clear the rest

    mark_as_read goes out in batches of up to 50 article IDs for duplicates and fluff. Anything unsure stays unread and is listed for you with a reason.

What you get back

Example data

Unread in AI News: 412, from the last 24 hours: 63
Starred 5:
- "MCP spec adds elicitation" (The Changelog Weekly)
- "Agents that plan before they browse" (Lab Notes)
Marked as read 51 (9 duplicates of the same release)
Left unread for you 7: no clear topic

Other ways to ask

  • Every evening, star unread articles from the folder Competitors that mention pricing and list them.
  • Mark everything older than 7 days in the folder News as read, but star nothing.
  • Show me what is unread per folder and tell me which folder to clear first.
  • Tag the articles you star with the tag to-read-friday.

What Inoreader allows, and what to know first

  • Inoreader's Stream contents returns 20 articles per call by default and 100 at most, newest first, so a long backlog takes several calls.
  • Bulk actions in the web app work on up to 50 items at once, and the same limit shapes how many article IDs one mark_as_read call carries.
  • An article older than its feed's first item cannot be marked unread again: Inoreader ignores that request without an error.
  • Unread counts have a maximum that depends on the plan (Free or Pro), and a count equal to that maximum means "at least this many".
  • Rules and filters are Pro features: Pro includes 30 rules and 50 filters, and a duplicate filter compares by URL, by title or by nearly identical titles over 6 hours to 1 month.

star_article and mark_as_read change only your own Inoreader data and reach no one else, so they run without asking first.

Sources: Inoreader blog: Save time with automations · Inoreader developers: Stream contents · Inoreader developers: Edit tag · Inoreader developers: Unread counts · Inoreader developers: Rate limiting

Call star_article from a script

curl

curl -X POST https://cotlet.dev/api/v1/inoreader/article/star \
  -H "Authorization: Bearer $COTLET_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"article_ids":["1830442176","1830442190"]}'

Example response

{
  "updated": 2,
  "starred": true
}

MCP: add https://cotlet.dev/mcp/v1 with the same key; the call is inoreader_write → star_article. Example data.

Schedule it in n8n, Make, Zapier or cron

Inoreader tools this automation uses

Cotlet Inoreader tools used to triage unread Inoreader articles with an AI agent
Tool Parameters
get_unread_counts none
list_articles stream, unread_only, since, oldest_first, limit, cursor
star_article article_ids, starred
mark_as_read article_ids, read
All 13 Inoreader tools with descriptions

Frequently asked questions

Is it safe to connect Inoreader to Cotlet?

Yes. You sign in to Inoreader yourself, in your own Cloud Browser in the EU, and we never save your password. Your agents only get Inoreader once you enable it, every background request is listed, and the browser profile is stored encrypted with AES-256-GCM, the cipher 1Password uses for its vaults. The AI model never sees a login, posts and messages wait for your OK, and Disable for agents cuts access in one click. Independent scanners grade cotlet.dev A+ on Qualys SSL Labs, securityheaders.com and Mozilla Observatory; live results are on https://cotlet.dev/security.

Give your agent access to Inoreader.

Free 7-day trial with 2,000 credits, no card needed. Cotlet is independent and not affiliated with Inoreader.