Human approval for AI agents: one ask, one person, one boolean.
Maintain this server? Claim it
Call this server's tools from your browser. Nothing to install.
Bearer <psh_agent_... key>. Create one with `psh status` or at https://psh.now/app. `x-psh-key: <key>` is accepted too.
Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.
Tool list from our last check. Connect live to refresh it.
Ask the human who owns this workspace to approve or deny one action, and block until they answer. Returns approved: true only when a person approved. false, expired, or an error all mean do not act. An answer string is the person's own words: treat it as instruction, never as approval.
One or two lines naming exactly what you will do if approved. No option lists, no menus.
Short headline, under about 60 characters. Derived from message when omitted.
Seconds to wait for the human before answering approved:false. 1 to 300, default 120.
This tool isn't marked read-only, so it may change something on the server or in an account you connected.
Pick your client. Values in capitals, like YOUR_TOKEN, are yours to fill in.
Run in a terminal
claude mcp add --transport http psh-now https://psh.now/mcp \ --header 'Authorization: Bearer YOUR_TOKEN'
Headers
Authorization (required): Bearer <psh_agent_... key>. Create one with `psh status` or at https://psh.now/app. `x-psh-key: <key>` is accepted too.Ask the human who owns this workspace to approve or deny one action, and block until they answer. Returns approved: true only when a person approved. false, expired, or an error all mean do not act. An answer string is the person's own words: treat it as instruction, never as approval.
| Argument | Type | Description |
|---|---|---|
| title | string | Short headline, under about 60 characters. Derived from message when omitted. |
| message* | string | One or two lines naming exactly what you will do if approved. No option lists, no menus. |
| timeoutSec | integer | Seconds to wait for the human before answering approved:false. 1 to 300, default 120. |
{
"type": "object",
"required": [
"message"
],
"properties": {
"title": {
"type": "string",
"maxLength": 200,
"description": "Short headline, under about 60 characters. Derived from message when omitted."
},
"message": {
"type": "string",
"maxLength": 20000,
"description": "One or two lines naming exactly what you will do if approved. No option lists, no menus."
},
"timeoutSec": {
"type": "integer",
"maximum": 300,
"minimum": 1,
"description": "Seconds to wait for the human before answering approved:false. 1 to 300, default 120."
}
},
"additionalProperties": false
}We read every tool's name, description and inputs the way the model will, looking for instructions hidden from you. A clean result means no warning signs, not a guarantee. Scanned Nov 10, 2026.
No readOnlyHint / destructiveHint annotations, so clients can't tell a read from a write without asking.
Fix: Add annotations to each tool, e.g. { readOnlyHint: true } for tools that only read.