Tempest AI

Run parallel coding agents with 64% fewer tokens

[Tempest AI @usetempest]

and you wonder why I exist 😭

[Prharsha G @gsvprharsha]

3.3B tokens in 30 days. On @ClaudeDevs pro plan!

Oh, I never let tokens go to waste.

So I never run out of limits.

@usetempest takes care of it. 🌪️

[Tempest AI @usetempest]

This is exactly why we chose to drive the actual CLI instead of reimplementing it.

When the upstream tool gets better, Tempest gets better too.

[Prharsha G @gsvprharsha]

Day One support for Claude Opus 5. 🚀

Tempest drives the real Claude Code CLI through a real PTY (xterm.js), so the moment Anthropic ships a new model, you get it.

No shims. No wrappers. No waiting.

Download Tempest for free now!
https://t.co/WDigFeDzOZ

[Prharsha G @gsvprharsha]

I never really shared why I'm building @usetempest.

So here it is.

I'm Prharsha, and over the past few years I've built products, shipped them, spent thousands of hours working with AI coding agents, and watched those tools evolve at an incredible pace.

One thing has become clear to me.

The future isn't a single AI replacing developers.

It's agentic engineering team with a human orchestrator.

AI agents are exceptional at executing well-defined tasks. They can research, write code, review changes, and test ideas in parallel.

But they still lack the judgment to decide what should be built, when priorities change, or whether the solution actually solves the problem.

That's where humans stay in the loop.

I believe the best engineering teams of the future won't be one developer and one AI.

They'll be one developer orchestrating a team of specialized agents.

But today's experience isn't there yet.

As you add more agents, token costs explode. Context gets duplicated. Every agent spends time rediscovering the same repository instead of building.

I'm building Tempest to change that.

Not just by making agentic engineering more efficient, but by making it safe, predictable, and easy to use.

Every feature we build is centered around one principle:

AI agents should work with you, not against you.

Humans should stay in control. Agents should handle execution. The collaboration should feel transparent, reliable, and intentional....not like handing your codebase to a black box.

I believe that's how software will be built over the next decade.

And that's the future I'm trying to build.

#buildinpublic

[Tempest AI @usetempest]

Small releases.
Fast iteration.
That's the plan.

[Tempest AI @usetempest]

v0.1.0 is out! Try now!!

[Prharsha G @gsvprharsha]

Building @usetempest has become my favourite part of the day.

v0.1.0 is out.

Last 48 hours:
+25,545 lines
−1,001 lines
4 commits

Still a long way to go.

On to v0.1.1

[Jack @jackcoder0]

🚨 Anthropic just showed a 24-minute workshop on how to actually do prompts for Claude.

Taught by the people who built it.

Free. No registration. No paywall.

I've seen $300 courses that don't cover what they teach in the first 8 minutes.

Watch it and bookmark it now.