Documentation
Start with what you want to accomplish.
TAI is a self-hosted runtime for agents that use tools, retain context, and keep working after you close the terminal. Pick an outcome below; the details appear when you need them.
Common paths
Get to a working result
01
Run your first agent
Install TAI and turn a repository's recent history into a useful standup.
Follow the quick start 02Connect a model
Use a local OpenAI-compatible server or install a provider for a hosted API.
Configure a provider 03Build a focused agent
Choose its instructions, tools, model, working directory, and operating boundaries.
Define an agent 04Automate recurring work
Choose a fixed cron job, an event-driven workflow, or a schedule the agent controls.
Compare scheduling optionsBrowse by area
Understand the system a layer at a time
Build agentsBehavior, tools, memory, tasks, and projects.Automate workWorkflows, cron jobs, hooks, and autonomous schedules.Connect & integrateChannels, shared rooms, and multi-agent conversations.Run & deploySelf-hosting, authentication, containers, and deploy targets.Extend TAICustom tools, MCP, skills, plugins, and TypeScript APIs.