Kova
Personal AI assistant and control-plane experiment for multi-channel, always-on workflows across terminal and messaging surfaces.
Chirag Borse — AI Systems Builder
I build modern AI systems, orchestration runtimes, developer tooling, and terminal-native workflows. My work sits around AI agents, MCP tooling, persistent memory, and local-first infrastructure.

┌───────────────────────┐ │ ██████╗███████╗ │ │ ██╔════╝██╔════╝ │ │ ██║ █████╗ │ │ ██║ ██╔══╝ │ │ ╚██████╗██║ │ │ ╚═════╝╚═╝ │ │ │ │ > agents: active │ │ > status: building │ └───────────────────────┘┌─────────────────────────────────────┐ │ │ │ ██████╗ ██████╗ ██████╗ ███████╗ │ │ ██╔════╝██╔═══██╗██╔══██╗██╔════╝ │ │ ██║ ██║ ██║██║ ██║█████╗ │ │ ██║ ██║ ██║██║ ██║██╔══╝ │ │ ╚██████╗╚██████╔╝██████╔╝███████╗ │ │ ╚═════╝ ╚═════╝ ╚═════╝ ╚══════╝ │ │ Chirag │ │ │ │ > agent systems: active │ │ > mcp tooling: shipping │ │ > memory graphs: online │ │ │ └─────────────────────────────────────┘
Systems
Personal AI assistant and control-plane experiment for multi-channel, always-on workflows across terminal and messaging surfaces.
Kova-compatible marketplace for publishing, discovering, and installing plugins, bundle plugins, and skills.
Graph-RAG powered personal AI memory system combining Neo4j, pgvector, entity relationships, and hybrid retrieval.
Security-focused Python CLI that scans MCP and AI-agent configs for leaked API keys, secrets, and credentials before they reach GitHub.
Local-first dashboard and API for tracing AI-agent runs, inspecting failures, tracking latency, and debugging workflow reliability.
Progressive web app for creators to save ideas, organize references, track tasks, and manage content workflows from one dashboard.
Self-hosted customer support chatbot built from scratch with LLaMA 3, FastAPI, session memory, and structured knowledge retrieval.
Current Focus
The areas I keep testing, breaking, and rebuilding while learning how practical AI systems should work.
Designing workflows where agents can coordinate, hand off state, and recover from tool failures.
Building command-line experiences that make AI systems feel local, fast, inspectable, and scriptable.
Exploring graph memory, hybrid retrieval, entity relationships, and long-lived context for AI assistants.
Shipping security and workflow tools around MCP servers, agent configs, and local AI development stacks.
Work in Progress
Active AI systems, local-first tools, and infrastructure experiments from my public GitHub.
Personal AI assistant, local control plane, and multi-channel agent runtime.
SQLite-backed reliability debugger for AI-agent traces, alerts, latency, and spend.
CLI safety layer for detecting secrets in MCP and AI-agent configuration files.
Hybrid graph/vector retrieval system for persistent personal AI memory.