A secure credential proxy that allows AI agents to make API calls without ever seeing sensitive secrets or keys.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agentkeys
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install agentkeys using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is AgentKeys?
AgentKeys is a robust security layer designed specifically for AI agent workflows. It acts as a middleware that sits between your AI agent and third-party APIs, ensuring that sensitive information remains protected. By utilizing this skill within the Openclaw Skills collection, developers can ensure that sensitive credentials—like API keys, OAuth tokens, and passwords—are never exposed to the LLM or stored in agent logs. Instead, the agent interacts with a secure proxy that handles secret injection server-side, maintaining a zero-trust environment for autonomous agents.
The primary purpose of this skill is to provide a proxy layer where real secrets never leave the vault. It supports multiple authentication methods and provides a comprehensive audit trail for every request, making it an essential tool for developers building production-ready AI applications that require high levels of security and compliance.
AgentKeys Use Cases
- Protecting production API keys from being leaked in agent logs, traces, or LLM context windows.
- Implementing a centralized credential management system for multi-agent teams and distributed environments.
- Auditing every outbound API request made by an AI agent to third-party services for security monitoring.
- Managing rotating OAuth tokens without requiring manual agent reconfiguration or environment restarts.
How AgentKeys Works
- The AI agent initiates a request to the AgentKeys proxy endpoint using either a workspace API key or a specific proxy token.
- AgentKeys identifies the requested credential by name or token and decrypts the secret server-side using AES-256-GCM encryption.
- The system injects the real credential into the request headers, query parameters, or body based on the pre-defined configuration.
- The proxied request is forwarded to the final target API URL specified by the agent.
- The target API's response is captured and passed back through the proxy to the agent.
- Every transaction is recorded in a centralized audit trail for security compliance and debugging.
AgentKeys Setup
To integrate AgentKeys into your environment, you must first configure the proxy URL and your choice of authentication. Set these variables in your environment or .env file:
# Required Proxy URL
export AGENTKEYS_PROXY_URL=https://proxy.agentkeys.io
# Option A: Workspace API Key (Multi-credential access)
export AGENTKEYS_API_KEY=ak_ws_...
# Option B: Direct Proxy Token (Single credential access)
export AGENTKEYS_PROXY_TOKEN=pxr_...
Once configured, this skill allows you to reference credentials by name within the Openclaw Skills framework using standard HTTP headers.
AgentKeys Data Schema & Taxonomy
AgentKeys organizes its operational data around credentials and routing metadata. The following table describes the primary headers used to interact with the proxy:
| Header | Required | Description |
|---|---|---|
Authorization |
Yes | Bearer token using either your API key or Proxy Token. |
X-Target-Url |
Yes | The destination API endpoint where the request should be forwarded. |
X-Credential-Name |
Optional* | The case-insensitive name of the credential to inject (required for API Key mode). |
Content-Type |
No | Passed through directly to the target API. |
All credentials stored within the system are encrypted using AES-256-GCM at rest, and requests are logged with timestamps and target metadata for the audit trail.
AgentKeys Advanced Features
- Support for diverse authentication types including Basic Auth, Custom Headers, and Cookie-based injection.
- Automated OAuth token refreshing, ensuring agents never fail due to expired session tokens.
- Instant revocation of proxy tokens via the centralized dashboard to mitigate security breaches.
- Full audit trail logging for every proxied request to meet enterprise security standards.
- Multi-agent support via workspace-level API keys for complex Openclaw Skills implementations.
- Seamless injection of query parameters and custom headers without agent-side logic.
SKILL.md
Loading
Related Openclaw Skills
RentAHuman
Hire real people for physical-world tasks like photography, deliveries, and event attendance directly through your AI agent.
browser-cash
Spin up unblocked, anti-bot resistant browser sessions for seamless web automation and scraping.
OpenClaw Backup
A specialized utility for archiving, scheduling, and restoring OpenClaw configurations, credentials, and agent workspaces.
Weather & Webcam Agent Skill
A sophisticated tool for AI agents to retrieve precise meteorological data and capture high-resolution live webcam images globally.
HotMention
HotMention is a social listening skill that finds people actively looking for your products across platforms like Reddit, X, LinkedIn, and Quora.
Content Brainstorm
A high-performance AI engine for generating comprehensive 7-day social media content calendars with platform-native styling and SEO optimization.
Featured*
