Documentation

Welcome to the Tailored AI documentation. TAI is a lightweight, modular AI agent framework optimized for local LLMs while supporting cloud providers like OpenAI and Anthropic.

Why TAI?

• Local-first: Designed from scratch to work with smaller models (30B parameter quantized models on consumer GPUs) by keeping system prompts short, tool counts low, and context tight.
• Hot-reloadable: Configuration, tools, and providers reload at runtime — the agent can modify its own capabilities without restarting.
• Simple by design: No complex state machines. The agent loop is straightforward: chat → tool calls → chat → stop.

What's in the box

• 3 providers: Ollama, OpenAI (+ compatible APIs), Anthropic
• 18+ built-in tools: Shell, file I/O, web search, browser automation, Gmail, Calendar, Drive, project tasks, and more
• Agent profiles: Named configurations with model/tool/behavior overrides and sub-agent delegation
• Multi-channel: CLI REPL, Discord bot, HTTP API with SSE, webhooks, cron scheduler
• Hooks system: Run tool calls before/after the agent loop for automation pipelines
• Custom tools: Define shell command tools in YAML config — no code needed

Get started

Head to Getting Started to install and run your first agent, or explore the sidebar to learn about specific features.