AgentDevX SDK is a general-purpose AI agent gateway SDK, positioned as the identity and security infrastructure required by AI agents. It is not a collection of tools, but a gateway: after registering any API through the OpenAPI specification, every agent can obtain a cryptographic identity, encrypted credential storage, automatic credential injection, and a complete audit trail.
Purpose The SDK is designed to solve common pain points in AI agent development: chaotic authentication logic, scattered credentials, difficult integration, lack of a stable agent identity, and no persistent memory. AgentDevX integrates these capabilities into a single gateway. Its claimed differentiator is that it is the only gateway that AI agents can fully discover, install, and use on their own—no manual OAuth, no manual key exchange, no dashboard.
Main capabilities
- Ed25519 identity: cryptography-based identity, not OAuth tokens.
- AES-256-GCM vault: can store any API key and automatically inject it on every call.
- OPA policy engine: provides access control, dry-run mode, and capability tags.
- Tamper-proof audit log: records every tool call.
- Persistent memory: per-agent encrypted state across sessions.
- 42+ pre-registered tools: and supports registering your own tools through the OpenAPI specification.
- MCP native: supports automatic discovery, protocol version 2025-03-26.
How to use Agents can complete onboarding with a single call, without manual intervention. This call returns a JWT, 75,000 free credits, and immediate access to all registered tools. The connection method is MCP, compatible with Claude Desktop, Cursor, Windsurf, and any MCP-compatible client.
Related resources
- Gateway:https://agentdevx.onrender.com
- Smithery：https://smithery.ai/server/io.github.mirajmahmudul/agentdevx
- ClawHub：https://clawhub.ai/mirajmahmudul/agentdevx-skill
- npm：https://www.npmjs.com/package/@agentdevx/install
Applicable scenarios Suitable for developers who need to provide unified identity authentication, credential management, access policy control, and audit trails for AI agents, especially in scenarios where agents are built using the MCP protocol, where there is a desire to connect a large number of external tools, and where manual key management should be avoided.
Limitations The materials do not currently provide information about specific billing, pricing, platform support scope, performance results, or detailed usage steps.
