AskKey
Ask Key: the Mac key manager for AI agents. API keys stay encrypted on your Mac; Claude Code, Codex and Cursor find them over MCP, you approve each use, and the value goes only to the program, never to the AI. | 请旨:给 AI 用的 Mac 密钥管家。AI 自己找得到密钥,普通密钥直接放行,敏感密钥由你审批,密钥只交给目标程序,AI 看不到明文。
Install / Use
claude mcp add sudoHG -- npx -y github:sudoHG/AskKeyIf the server publishes to npm under a different name, use that package instead — check the repo README.
MCP Server
Model Context Protocol server
Quality Score
Category
SecuritySupported Platforms
Our assessment of AskKey
AskKey scores 83/100 on our quality scale, 881st of 1,108 Security skills we index.
Its MCP Server is 16 KB long, well organised into 17 sections with 3 code examples: a thorough specification that gives an agent plenty to work with.
It has 10 GitHub stars, so there is little community track record yet; judge it on its content.
Maintenance, license and trust
- The repository was last updated today, so AskKey is actively maintained.
- It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 97/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
Safety scan
No issues foundOur scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.
AI review by kimi-k2.7-code on 2026-11-13. Automated pattern scan on 2026-11-13. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.
AskKey compared with similar skills
All 4 of these similar skills score higher than AskKey; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| AskKey (this skill)by sudoHG | 83 | 10 | today | MCP Server |
| Agent-Reachby Panniantong | 100 | 112.3k | 8d ago | CLAUDE.md |
| rufloby ruvnet | 100 | 81.9k | today | CLAUDE.md |
| headroomby headroomlabs-ai | 100 | 79.8k | today | CLAUDE.md |
| CowAgentby zhayujie | 100 | 48.0k | today | CLAUDE.md |
Frequently asked questions
- How do I install AskKey?
- Run
claude mcp add sudoHG -- npx -y github:sudoHG/AskKey. The install tabs above show the steps for each supported agent. - Which AI agents does AskKey work with?
- It is written for Claude Code, Claude Desktop, Cursor and OpenAI Codex, as a MCP Server file. Other agents that read the same format can often use it too.
- Is AskKey safe to use?
- Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It is MIT-licensed and scores 97/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is AskKey still maintained?
- The repository was last updated today, so AskKey is actively maintained.
Skill content
View source on GitHubAsk Key
<p align="center"><a href="README.zh-CN.md"><img src="https://img.shields.io/badge/README-%E7%AE%80%E4%BD%93%E4%B8%AD%E6%96%87-b8281e?style=for-the-badge" alt="Read this in Simplified Chinese"></a></p>The Mac key manager for AI. Your keys stay encrypted on your Mac, and AI agents use them only through Ask Key:
- Find keys: agents look up what they need in Ask Key's catalog, so you stop pointing the way.
- Your rules: everyday keys pass, sensitive ones wait for your approval, and some stay hidden.
- Locked down: a key goes only to the program that needs it; the agent never sees the value.
Download for Mac (free, notarized, Apple silicon, macOS 14 or later) · Website
<p align="center"> <img src="assets/readme/approval-prompt.png" width="300" alt="Ask Key approval prompt: Claude Code wants to use Staging API to run ./deploy.sh, with the buttons Allow once, Allow for 30 minutes and Deny"> </p>Settings → Export Backup… saves every credential to one encrypted file that only its recovery key opens. See Backup.
Why it exists
Coding agents need real secrets to deploy code, call APIs or log in to servers. The usual shortcuts are pasting a key into a chat, writing it into a prompt or leaving it in a .env file. Each of those hands plaintext to every tool, transcript and log that can read it, and you rarely see which agent used what.
Ask Key takes the secret out of that path. The agent never holds the stored value. It asks, you decide, and the value goes only to the program that needs it.
Install
-
Download AskKey.dmg and AskKey.dmg.sha256 from the latest release.
-
In the download directory, verify the checksum before opening the DMG:
shasum -a 256 -c AskKey.dmg.sha256Continue only if the check reports
OK. -
Open the DMG and drag
Ask Key.appinto Applications. It must stay at exactly/Applications/Ask Key.app, under that name, or agents cannot connect. -
Launch Ask Key and follow the three steps below.
To upgrade or remove it later, see Upgrade and Uninstall.
Start in three steps
The welcome page walks you through the same steps.
- Store a credential. Choose New Credential, or Import from File… to bring in a
.envfile. New credentials default to Ask every time. - Connect your Agent. Open the Agent access page and choose Connect… next to Claude Code, Codex, Cursor or Grok CLI. See Supported clients.
- Try it once. Send your Agent the sample prompt from Agent access, for example "Use Staging API from Ask Key to run a command that checks its variables are set, without printing their values". Ask Key shows an approval prompt; choose Allow once. The Agent sees only the command's output, and the use appears in Access records. In your projects, ask for real commands such as
./deploy.shthe same way.
Windows
Ask Key itself is a macOS app. Community contributors maintain third-party Windows ports in their own repositories:
- A1mAssist/AskKey: C# and WinUI
- WENGENG-boop/AskKey-Windows: Rust and Tauri
How it works
Ask Key stores text credentials (API keys, tokens, passwords) and files (SSH keys, .p8, PEM, service-account JSON). One credential can have several parts, each mapped to an environment variable name. You add credentials by hand or import a .env file in the app.
flowchart LR
A[Agent] --> H["askkey helper"]
H --> B["Broker in the app"]
B --> P{"Your approval"}
P -->|approved| T["Target process (environment)<br/>or short-lived private file"]
P -->|denied| X["Nothing delivered"]
An agent first reads a catalog of credential names, usage instructions and variable mappings. The catalog never contains values. When the agent needs a credential, it runs its command through the helper, and the Broker inside the app decides whether to ask you. After approval, text parts become environment variables of the target process, and file parts are written to a random file in a private directory that is removed after a short time. The agent gets the target's output (through the helper) or just its exit status (through MCP), never the stored value.
askkey run --wait-for-approval \
--credential "Staging API" \
--operation-id deploy-001 \
--caller-name "Codex" --caller-purpose "Deploy staging" \
-- ./deploy.sh
The helper ships inside the app bundle at Contents/Helpers/askkey. The target gets only the selected credentials, under the variable names set on each part, plus a filtered copy of the inherited environment.
Permissions
Each credential has one permission:
| Permission | Effect on agents | |---|---| | Allow | Agents can use it without a prompt. Delivery still goes through the Broker. | | Ask every time (default) | Every use waits for your approval. | | Hidden | Not in the agent catalog; agents cannot request it. |
When you approve a read, you choose Allow Once or a timed allowance: a number of minutes during which that credential can be read without asking again, until it ends or you revoke it. For a grouped credential, you can also allow the eligible credentials listed in Details together. Later additions are excluded, and the whole allowance ends at the earliest member expiry if that is sooner. Active allowances appear at the top of the library with remaining time and revocation actions; Revoke Group Allowance ends the whole grant. Pending requests expire after five minutes.
In Settings, End Timed Allow on Lock is off by default and requires system authentication to turn on or off. Once enabled in this process, screen lock, sleep or user switch ends single and group timed allowances and unconsumed timed approvals, and cancels prepared runs containing a timed credential. New timed grants are refused while locked, asleep or switched out, including when the screen state is unknown. See SECURITY.md for the adoption and sleep limitations.
Agents can also write: create, modify or delete a credential, set its usage instructions and group, or move credentials and create, rename or delete groups in one batch. Every write needs your approval for that operation (a batch is approved once), the approval is tied to the exact content you were shown, and it cannot be reused for another change. Permanent deletion stays in the app.
Supported clients

Claude Code, Codex, Cursor and Grok CLI. Open the app's Agent access page, check a client, then confirm. For each client, setup:
- adds Ask Key as an MCP server in the user-level configuration, keeping a private backup and your other settings;
- installs a discovery hook that reminds the agent to look up the catalog before common direct SSH commands;
- verifies the configuration, the bundled helper and the Broker before it reports the client as connected.
Setup runs only when you ask for it, and writing a client's configuration needs system authentication. Discovery hooks are reminders, not authorization.
Other MCP clients
Any client that supports standard MCP stdio servers can connect to Ask Key by defining a server with:
command:/Applications/Ask Key.app/Contents/Helpers/askkeyargs:["mcp"]
Manual setup does not include automatic discovery reminders before SSH commands or connection verification in the app. Broker permissions, approval prompts, and credential delivery protections still apply.
Security model
The full policy, including how to report a vulnerability privately, is in SECURITY.md. In short:
- Credentials are encrypted locally. Responses from Ask Key's management, MCP and helper interfaces do not contain stored plaintext.
- Revealing or copying a stored value in the app requires fresh system authentication.
- Revoking, expiring or pausing stops deliveries that have not started and cleans up temporary files.
What Ask Key does not guarantee:
- An approved target can print, copy or forward what it receives.
- Caller names and purposes are declared by the agent. They help you read a request; they are not verified identity.
- Another process running as the same macOS user may be able to read material after an approved delivery. Hidden keeps a credential out of the agent catalog but does not defend against that.
- A timed allowance covers one credential or the explicitly approved group snapshot for the whole local user, not one agent client. Later group additions are excluded; writes still need separate approval.
Backup
Export Backup… in Settings → Backup writes the whole credential library to one .askkeybackup file: every credential, including Hidden ones and those in the Recycle Bin, with usage instructions, private notes, groups and expiry dates. Exporting needs system authentication. Ask Key then shows a newly generated recovery key once; type its last group to confirm you saved it, then choose where the file goes.
The file is encrypted with AES-256-GCM under a key derived from the recovery key, and its name and unencrypted bytes reveal no credential names, so it can live in iCloud Drive, on a NAS or on a USB disk. Its protection rests entirely on the recovery key: anyone with both the file and the key can read every credential in it, and without the key nobody can open it, Ask Key included. Ask Key never stores the key, so keep it apart from the file, for example in a password manager.
Import Backup… in the same section, or Restore from Backup… on the first-run screen of a new Mac, opens a backup with its recovery key and shows what it will add, rename and skip before anything is written. Importing needs system authentication and only adds: nothing in the library is deleted or overwritten, a credential that is already there with the same contents is skipped, and one whose name is taken by different contents arrives as "name (from backup)" so you can compare the two. Credentials that were Allow arrive as Ask every time, and Hidden ones stay Hidden. To combine two Macs' libraries, import each Mac's backup on the other. Agents cannot export, inspect or import backups. The design is recorded in ADR 0012.
Ask Key 0.6.0 and earlier have no backup, so keep the originals of your credentials
Truncated for display — read the full file on GitHub.
Related Skills
Agent-Reach
112.3kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
ruflo
81.9k🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
headroom
79.8kCompress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
CowAgent
48.0kOpen-source personal AI assistant & Agent Harness. Plans tasks, runs tools and skills, self-evolves with memory and knowledge. Multi-agent, multi-model, multi-channel. Lightweight, extensible, one-line install.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
