In this entry of my AI DX/Tools series, I want to discuss Amp: an AI coding agent designed for developers who spend most of their time in the terminal and editor, rather than in chat windows.

Amp is free to try, works locally, and integrates directly with your existing workflow. It doesn’t try to impress with flashy UI or novelty features. Instead, it focuses on making AI usable for real engineering work.

---

Opinionated by Default (and That’s a Good Thing)

One of the strongest characteristics of Amp is that it is very opinionated.

You don’t manually choose models, tweak parameters, or debate which AI is best for which task. You choose intent, and Amp chooses the model for you. This removes a surprising amount of friction. Instead of thinking about tooling, you stay focused on solving the problem.

This design choice won’t appeal to everyone, but for daily work, it’s refreshing.

---

Three Modes That Match How Developers Think

Amp offers three modes, each mapping cleanly to real-world usage:

• Smart – Unconstrained, state-of-the-art model usage for complex reasoning and bigger changes.

• Rush – Faster and cheaper, suitable for small, well-defined tasks.

• Free – Free to use, backed by fast basic models, with subtle ads shown in the terminal.

Instead of asking “Which model should I use?”, you ask “How complex is this task?” That’s a much better abstraction.

---

Handoffs: A Practical Solution to Context Limits

Anyone who uses AI seriously eventually runs into context problems. Threads get noisy. Errors pile up. Important details fall out of scope.

Amp’s /handoff command is one of its most practical ideas. It allows you to intentionally move work from one thread into a new one, carrying only the relevant context.

Rather than fighting context windows or relying on aggressive summarization, you reset cleanly and continue. For longer or multi-phase tasks, this makes a huge difference.

---

Threads Are Shareable Artifacts

Another highlight is thread sharing.

You can share a conversation with teammates or make it public, allowing it to appear on your profile. Anyone with the link can see the full history and understand the reasoning behind decisions.

For teams, this changes how AI is used. Conversations become something you can reference, review, and learn from, much like pull requests or design docs.

---

Free, Accessible, and CLI-First

Personally, I really like Amp CLI in free mode.

It’s extremely accessible. You install it, sign in, and start using it. The ads are subtle and blend quietly into the terminal without disrupting your workflow. Although it is free, the tool remains genuinely useful.

This makes Amp easy to recommend to teammates or friends who are curious about AI-assisted development but don’t want to commit upfront.

---

Final Thoughts

Amp doesn’t try to be clever or flashy. Its value comes from being:

• Opinionated

• Workflow-aware

• Designed around real developer problems

If you care about AI as part of your daily engineering workflow, not just as a demo or experiment, Amp is worth trying, especially on the free tier.

In future posts, I’ll likely compare Amp with other agent-based tools and explore where this style of AI-assisted development is heading.