See which agents are running, what they're doing, and how deep the delegation goes — in real time.
Track token usage per agent. Know which sub-agents are expensive and which are efficient.
No other terminal AI interface visualizes agent orchestration. This is unique to Amplifier TUI.
Every AI tool user asks "why is it forgetting things?" This answers that question visually, in real-time. See your context window fill up, know when you're running low, and act before it's too late.
Create a branch from any point in the conversation to explore a different direction.
Jump between branches instantly. Compare responses side by side to pick the best approach.
Found what you want? Merge it back to your main conversation thread.
Send the same prompt to Claude and GPT-4 at the same time. Watch both responses stream in.
Compare quality, speed, and style. Choose the response you prefer and continue with that model.
Running model evaluations? A/B testing makes it effortless to compare outputs on real prompts.
Works with any provider Amplifier supports — Anthropic, OpenAI, Azure, local models, and more.
0.5x for studying, 2x for review, instant for jumping to the end. Full playback control.
Jump to any point in the session. See messages appear with their original timing.
Show stakeholders how a complex agent session unfolded. Better than screenshots.
Review past sessions to understand what worked, what didn't, and how to prompt better.
62 hours, 9 prompts, 147 commits. One autonomous session built the entire TUI from scratch.
14 features overnight, 659 new tests. A structured feature backlog executed autonomously.
922 tests, 0 errors, 2 sessions. A production-quality terminal UI built almost entirely by AI agents.
Data as of: February 2026
Feature status: Active
Research performed:
gh repo view ramparte/amplifier-tuipytest --collect-only (922 tests collected)Gaps: Exact hour count for marathon session is approximate based on session timestamps; commit counts from git log
Primary contributors: ramparte (sole author, ~100% commits)