Tools
Paperzilla Openclaw
By Paperzilla OAuth MCP plugin for OpenClaw: morning research briefs, public digests, papers and owned-feed feedback.
Install
openclaw plugins install @pors/paperzilla-mcpREADME
# Paperzilla MCP for OpenClaw
Start your morning with new research worth reading. Paperzilla supplies focused
feeds of scientific papers. This plugin combines one or more feeds into a short
brief with source links and an explanation of why each paper fits your interests.
It includes the production OAuth MCP connection and a Morning Brief skill.
It connects directly to Paperzilla; the `pz` command-line tool is not required.
The existing [Paperzilla CLI skills](https://github.com/paperzilla-ai/paperzilla-skills)
remain available separately.
## Install and connect
Requires OpenClaw **2026.10.8 or later**. Use the Gateway's own HTTPS address,
Tailscale address, or local loopback address for browser sign-in.
1. In OpenClaw's **Plugins** catalog, find **Paperzilla MCP** by `pors` and install
it. The package is `@pors/paperzilla-mcp`; check that identity before installing.
2. Open **Plugins → Paperzilla MCP → Overview → Accounts** and select **Connect**
for `paperzilla`.
3. In the browser, sign in to Paperzilla or create an account. Check the account
and permissions on the consent screen, then select **Allow**.
4. Return to OpenClaw and start a fresh chat. Ask it to check your Paperzilla
connection and list your feeds.
The equivalent CLI installation is:
```sh
openclaw plugins install @pors/paperzilla-mcp
```
The plugin supplies the server definition. A bundle connection can appear under
the plugin's **Accounts** without appearing in **Settings → MCP**, which lists
manually configured servers. Use the plugin's **Connect** action for sign-in.
You do not need an API key or a second server entry.
An existing manual `paperzilla` server definition takes precedence over the
plugin's default. If you already configured Paperzilla, check its URL and scopes
before connecting; a beta definition must not be mistaken for production.
Installing the plugin does not remove or replace your manual configuration.
## Try a morning brief
> Help me set up a daily morning research brief from one or more of my Paperzilla
> feeds. First show my feeds so I can choose them. I care about reliable AI agents
> and evaluation. Suggest three to five papers, with links and reasons to read.
Choose feeds, a time and a timezone. The skill checks fresh additions, combines
duplicate papers across feeds, and reports the feeds and time window covered.
Scheduling uses OpenClaw's available automation tools and your authorization;
the plugin itself does not create a schedule during installation. Try a manual
run before relying on unattended delivery.
For smoother unattended briefs, allow Paperzilla's read-only tools without
repeated prompts where your OpenClaw approval settings support it. Keep Markdown
generation and feedback writes approval-required unless you choose otherwise.
OAuth consent and tool-call approval are separate controls.
## Explore before creating a feed
A Paperzilla account is required. Every account can use public digests and papers
already in the catalog; a paid subscription or project is not required for those
tools. To build a brief from your own feeds, [create a project](https://paperzilla.ai/digests/new).
> Find public Paperzilla digests about machine learning, open the latest issue
> of a relevant digest, and show three papers with source links.
> Read prepared Markdown for this Paperzilla paper and explain its methods,
> findings and limitations. Tell me if you could only read part of it.
The integration can look up Paperzilla IDs and URLs, DOIs, and supported source
URLs against the existing catalog. It does not import missing papers or create
projects.
## Permissions and writes
The browser asks for these permissions:
| Permission | Purpose |
| --- | --- |
| `catalog:read` | Public digests, catalog paper details and prepared Markdown |
| `projects:read` | All current and future projects you own, including feeds |
| `feedback:write` | Explicitly save or clear feedback on papers in your own feeds |
| `markdown:generate` | Explicitly request Markdown preparation for an existing paper |
| `offline_access` | Refresh the connection for later use |
A morning brief only reads research. It does not save ratings or start Markdown
processing automatically. Markdown reads never enqueue work. Explicit generation
requests are limited to ten per minute per account across clients; source
availability, deduplication and failure cooldowns still apply.
To save feedback, name the paper and its owned feed, for example:
> Star the first paper from that brief in my Reliable AI Agents feed.
A vote replaces an earlier vote, including a star. Ask to clear feedback to remove
it. Public-digest access does not grant permission to rate another person's feed.
## Accounts, revocation and troubleshooting
An OpenClaw Gateway can share a Paperzilla connection across agents or people.
Connect the intended account and use OpenClaw's identity and tool-access controls
for shared deployments. The plugin does not isolate each Telegram sender into a
different Paperzilla account.
To disconnect, revoke the relevant client at
[Paperzilla → Connect AI clients](https://paperzilla.ai/ai-connections).
Uninstalling a plugin alone is not a guarantee that saved OAuth credentials are
revoked. Reconnect through the plugin's **Accounts → Connect** action when needed.
If tools are missing, check that the plugin is enabled and your agent can use
bundle MCP tools. A fresh connection check distinguishes missing tools from
authentication errors or an account without projects. See the
[OpenClaw MCP guide](https://docs.paperzilla.ai/guides/openclaw-mcp) for callbacks,
scope errors, manual-server overrides and additional diagnostics.
## Support and maintenance
- [Website](https://paperzilla.ai) · [AI workflows](https://paperzilla.ai/agents)
- [Setup guide](https://docs.paperzilla.ai/guides/openclaw-mcp)
- [Support](mailto:hello@paperzilla.ai) · [Issues](https://github.com/paperzilla-ai/paperzilla-openclaw-plugin/issues)
- [Privacy](https://paperzilla.ai/privacy) · [Terms](https://paperzilla.ai/terms)
Maintained by Paperzilla. This is a declarative plugin bundle: a remote MCP
definition, skill instructions and artwork. It has no executable runtime,
install scripts or local CLI dependency. The plugin's instructions are MIT
licensed; Paperzilla artwork and trademarks remain Paperzilla's property.
tools
Comments
Sign in to leave a comment