AI agent you install and point at problems
Zero-config AI agent with real shell access, file I/O, web browsing, and persistent memory. Bring your own key, use local models, or start free.
$ npx agntk "fix the failing tests" What it does
Zero-Config Providers
Bring your own API key, run local models with Ollama, or start free — no setup required. agntk auto-detects what's available.
Real Tools
Shell commands, file read/write, glob search, web browsing, and more. The agent works with your actual codebase, not a sandbox.
Named Agents & Memory
Create persistent agents that remember context across sessions. Each agent has its own identity, instructions, and memory store.
Hardware-Aware Local AI
Auto-detects your system — Apple Silicon, NVIDIA GPU, available RAM — and picks the right Ollama model for your hardware.
HTTP Server & Client
Serve agents over REST, SSE, or WebSocket. Connect remotely with the client SDK. Build agent-powered apps and services.
Modular SDK
Six focused packages — core, CLI, server, client, logger, and the npx wrapper. Use the pieces you need in your own projects.
Zero-config provider cascade
agntk checks for API keys, probes for a local Ollama server, and falls back to the free tier — all automatically. You don't configure anything.