Share a folder of papers with Claude Cowork; it extracts the finding and a figure from each, writes one X post of up to 280 characters and saves drafts in Tareno for your approval.
Can Claude Cowork read new papers and prepare one X post for each?
Yes. Claude Cowork reads the PDFs in a shared folder, extracts the main finding and the limits as the authors state them, writes one post per paper within 280 characters and saves drafts with the figure you choose through Tareno. You verify every number against the paper, and approval happens in Tareno.
Researchers and research-minded teams want to share new papers, but writing a post is the last thing anyone does after reading forty pages. The temptation is to let an assistant summarize and post, which is exactly where mistakes become public: a result stated more strongly than the authors wrote it, a number from the wrong table. Claude Cowork works on a folder of files, so it can read the papers themselves rather than a press release, and it can be told to quote the authors' own limits. This page follows a lab group through a weekly digest, with the checking steps that keep the posts honest.
Example: a robotics lab shares three new papers each week
Imagine a robotics lab whose communications student puts three new PDFs in a Cowork folder each Friday, together with the figure files exported by the authors. Cowork reads each paper, writes down the main finding, the sample size and the authors' own stated limitation, and drafts one post per paper under 280 characters. In the second paper it notices that the abstract reports a 12 percent improvement while the results table shows 9, and flags the difference instead of choosing one. The authors confirm the table. Cowork finds the figure in the library with list_media, saves the three drafts through create_drafts, and the student requests Friday 15:00 slots.
Step by step
01
Share the papers and the figures
Put the PDFs and the figure files the authors released in one Cowork task. A figure with a clear license is better than a screenshot of a page.
02
Extract the facts before any wording
Ask Cowork for a table per paper with the main finding, the sample, the number reported and the limitation the authors name. Nothing is written for X until the table is checked.
03
Compare abstract and results
Ask it to flag any number that differs between the abstract and the results section. Authors' abstracts are sometimes rounder than their tables.
04
Write one post per paper
Request a post within 280 characters per paper that states the finding as the authors did, and keeps the limitation if the space allows. If it does not, put the limitation in the linked paper summary.
05
Attach the figure and save
Cowork finds the figure with list_media and saves each post with its image through create_drafts for your X account. Up to four photos fit in one post.
06
Request times and approve
Give dates, times and an IANA time zone. The requests become pending actions, and you approve them in the Tareno dashboard.
Prompts to copy
Extract before writing
Read the three PDFs in the shared folder as source material for Tareno drafts. For each paper return a table with the main finding, the sample size, the headline number and the limitation the authors name, each with the page it comes from. Do not write any post until I confirm the table.
Find the mismatches
Compare the abstract with the results section in each paper and list every number that differs, with both values and both page numbers. Tell me which one you would trust and why, but leave the decision to me.
One post per paper
Write one X post per paper, for later saving in Tareno, within 280 characters that states the finding in the authors' own strength, without words like proves or breakthrough. Count the characters, keep the paper's link, and mention the limitation if it fits.
Match the figures and save
Use Tareno to find the figure files for these papers in my media library, save each approved post with its figure as a draft for my X account [handle], and return a table of draft ID, paper and figure. Request no time yet.
X limits and tools in Tareno
Text
up to 280 characters; hashtags and mentions are supported; plain text, no rich formatting
Images
jpg, png, gif, webp, up to 5 MB each, up to 4 per post, recommended ratios 16:9, 1:1
Video
mp4, mov, 0.5 to 2 min, up to 50 MB, one video per post, ratio 16:9, 1:1
Links
links can sit inside the text and get a preview card
What Tareno can read
posts published outside Tareno: no; comments: no; transcripts: no; video download: yes
As configured in Tareno on 2026-10-24. The live rules for your own account come from the get_platform_schema tool.
What X itself documents
A post may include up to 4 photos, 1 animated GIF or 1 video.
The allowed video length depends on the posting account, not on the API plan: 20 minutes and 8 GB by default, 125 minutes and 16 GB with X Premium or verified status. Tareno's configuration for X stops at 140 seconds.
Quote-posting through the API requires an Enterprise plan and is not available on self-serve tiers.
Tareno's configuration for X allows 280 characters and up to four photos, so a post can carry the finding and its main figure, but not a full results table.
X's API documentation allows up to four photos, one animated GIF or one video in a post, which means a short animation from the paper can replace a still image.
Tareno cannot read replies on X, so questions that researchers ask under the post never reach Cowork; someone has to read them on X.
X does not render formatting, so a result with a subscript or an exponent has to be written in plain characters, and Cowork should say when a symbol will not survive.
What does not work
Claude Cowork cannot judge whether a paper's method is sound, so a person with domain knowledge reads the extracted table before approval.
It cannot read the replies that researchers write on X, because Tareno does not expose replies for X.
It cannot publish a post on its own, since a time request becomes a post only after you approve it in the Tareno dashboard.
Mistakes to avoid
Strengthening a result. Words like proves, solves or breakthrough rarely match a paper's own claim, so ask Cowork to avoid them.
Using a figure without its license. Take figures the authors released for reuse, and credit them in the text.
Posting a preprint as if it were settled. Say in the post when a paper is not peer reviewed.
Picking the number from the abstract. Check it against the results section before the post goes out.
Open the published Tareno listing, choose install or connect, and complete the Tareno login. Enable Tareno in a new conversation. For Claude, search for Tareno in the supported plugin directory.
Choose OAuth, connect and sign in with your own Tareno account. Check the requested permissions. Refresh the client's tool list after authorization.
Run list_accounts first. Confirm the returned workspace and account IDs before creating a draft. If no accounts appear, connect a social account in Tareno first.
https://tareno.co/api/mcp
Use the same Tareno account across supported clients. Open the relevant setup guide for authentication, first use, reconnect, and revocation.
Does Cowork read the whole paper or just the abstract?
It reads the PDF in the shared folder, so it can compare the abstract with the results section. That is why we ask for a fact table with page numbers first: you can verify each number in the paper before a post exists.
How should a limitation appear in a 280-character post?
In a few words if it fits, for example by naming the sample. If it does not fit, link to a longer summary that states it. A post that leaves out a limitation the authors named overstates the work.
Can a post include the paper's key figure?
Yes. Tareno's configuration allows up to four photos on X, and Cowork can find the figure file in your library. Use figures the authors released for reuse and put the credit in the text.
Why does Cowork flag differences between abstract and table?
Because authors sometimes round or update the abstract. Cowork reports both values with page numbers and leaves the choice to you, which is safer than letting it pick the more flattering number.
Can Cowork answer researchers who reply to the post?
No. Tareno does not read replies on X, so Cowork never sees them, and no tool posts a reply. Someone on the team answers on X, using the fact table as a reference.
Does the weekly digest post itself?
No. Requests become pending actions in Tareno, and X receives the posts only after you approve them in the dashboard. `get_action_status` shows what happened to each one afterwards.