If you follow a research field, you probably have a Google Scholar alert or two. Maybe a dozen. Some from when you were starting your thesis. Some from a grant you wrote in 2022. Some from a keyword that was hot eighteen months ago and now shows up in half the papers you skim.
You know the feeling: a pile of emails with links, most of them only loosely relevant, no structure, no ranking, no memory of what you are actually working on this week.
There is a better way now.
What Paperzilla does
Paperzilla watches preprint servers and journals for you. You describe a research topic once (a thesis, a grant, a product area), and Paperzilla monitors arXiv, bioRxiv, medRxiv, ChemRxiv, PubMed and friends for new papers that match. Each paper gets a structured summary plus a weaknesses analysis, and everything lands in a feed per topic.
You can read that feed in the app, in an email digest, or through RSS. That is the human side.
What changed in the last few months is the other side: your AI agent can now read the same feed. Via CLI or MCP. Structured data, ready for automated consumption, and importantly, the full markdown of each paper so your agent can actually read the thing, not just the title.
From alert to brief
This is the shift. Instead of twenty loose emails from Scholar, you get one short brief from your own agent every morning.
Your agent knows your feed. It also knows what you are working on right now. Combine those, and it can pick the one paper out of today's batch that is worth your ten minutes, explain why, and warn you about the caveats.
Here is the kind of note my agent sends me each morning:
| 🦞 |
CLAWD Morning brief, Living Claims Ledger Agent Top paper today: AISSISTANT: Human-AI Collaborative Review and Perspective Research Workflows in Data Science Because you're currently exploring human-in-the-loop research workflows, this one is unusually close to the direction you care about. Why it matches:
Caveat: not a perfect product fit, it leans more toward review/perspective writing than continuous monitoring, but it still feels worth reading. |
Then we can keep discussing the paper, in the same place the brief came from. My agent already knows what I care about, so all the context stays in one spot.
If you want to see this happening in real life, I keep a public, read-only Telegram channel where you can watch my weekday brief arrive and watch me discuss it with the agent: t.me/paper_claw.
Pick your agent, pick your channel
There is no one way to do this. Paperzilla exposes both a CLI and an MCP endpoint, so most coding agents and chat agents can work with it. The channel is whatever your team already lives in.
Three setups with step-by-step guides today:
- OpenClaw + Telegram: open-source agent driving the
pzCLI, posting into a Telegram channel. Guide. - Claude Cowork + Slack: Claude wired up via Paperzilla MCP, posting into a Slack channel. Guide.
- Codex + Microsoft Teams: Codex with Paperzilla MCP, posting into a Teams channel. Guide.
The custom skills I use to make this work are in a public repo, so you can copy them instead of starting from scratch: github.com/paperzilla-ai/paperzilla-skills.
Bridging the old and the new
If you have Scholar alerts you rely on and don't want to rebuild a topic from zero, we are working on a bridge: forward your Scholar alert emails to a Paperzilla address, and we extract the underlying papers into a project feed. Your agent then reads from there as normal.
This is in alpha. If you want to be part of it, email me at mark@paperzilla.ai.
What's next
- Scholar alert forwarding (alpha): seed a project feed from alerts you already trust. Contact me to join the alpha.
Try it
Start a project: paperzilla.ai.
Watch a brief in action: t.me/paper_claw.
Questions? Email mark@paperzilla.ai.