Drakeify is a lightweight agent service that runs between your LLM and the applications that use it.
Out of the box, Drakeify behaves as a smart proxy. It sits in front of your LLM, normalizing model output and restructuring it into reliable, consistent responses for your clients.
Drakeify also provides a plugin and tool system that allows you to extend and customize how your LLM behaves.
Built as a stateless, horizontally scalable agent platform, with externalized session state, memory, and configuration to support distributed operation across many runtime nodes.
Drakeify is designed to support a wide range of use cases, from personal automation to large-scale enterprise deployments.
For individuals, Drakeify can power a personal assistant capable of scheduling tasks, interacting with services, sending messages, and responding to events.
For organizations, Drakeify provides a horizontally scalable agent runtime with centralized state management and configurable guardrails, allowing AI-driven workflows to interact safely with internal systems and APIs.
Key features:
-
Proxy normalization – Drakeify cleans up inconsistent LLM responses and converts them into reliable structured output for clients.
-
Plugins – Plugins can modify the conversation flow, allowing you to inject context, enforce rules, filter responses, or add automation.
-
Webhook bots – Plugins can expose webhooks, making it easy to build bots that integrate with chat systems like Zulip, Slack, or Discord.
-
Tools – Tools are small, self-contained plugins that give the LLM new capabilities (running commands, searching systems, sending messages, etc.).
-
Tool registry – A built-in registry allows tools and plugins to be easily shared and discovered.
-
Transparent proxying – Drakeify tools remain invisible to the client while still allowing client-side tools to pass through normally. Agents like copilot and claude can be used, with enhanced capabilities.
for plugin development, see PLUGINS.md.
for tool development, see TOOLS.md.
for implementation details, see IMPLEMENTATION.md.