For AI agents: the complete documentation index is available at https://agentia-web.pages.dev/llms.txt, and the full documentation bundle (the single-source usage guide, plain text, in Chinese) is available at https://agentia-web.pages.dev/llms-full.txt.

Simulated demo: a local pre-recorded script, not a real model call

Watch a run happen

Pick a task and watch the main agent think, pick a capability off its menu, open an llm.turn, call the capability and assemble the result. On the right is Agentia's observability surface itself: the trace call tree and the token / cost figures grow in step with the replay — one run == one trace, built in from turn 0, not a separate tracing SDK bolted on.

02 — REPLAY

run replay

Pick a task on the left and hit "Run" to start the replay