stagewise

Approving every tool call your agent makes gets exhausting fast. So we built Smart Approvals to fix that. There are two ways agents typically handle bash tool calls today: - "Always Ask" means the agent stops and waits for your confirmation on every single command — safe, but it breaks your flow constantly. - "Always Allow" means it runs everything without asking — fast, but there are no guardrails around dangerous actions. Smart Approvals uses a classifier that sits between those two extremes. It reads the previous agent actions, the current shell state, and the command it's about to run. Safe commands get auto-approved silently. Suspicious ones get flagged with the reason, so you can deny or approve with full context. You move faster without handing over full control.

Last week, we shipped six releases, focusing on much-requested features: ⚫️ GPT 5.5 support Once GPT 5.5 was available on the public API, we immediately shipped it as a new model in our app. OpenAI's latest frontier model comes with a new level of intelligence and agentic capabilities, but costs significantly more. If you need its new capabilities: The model is available in stagewise now! 👀 Agent Preview Panels Running multiple agents at once can quickly become a hassle. We shipped Agent Preview Panels to see an agent's work at a single glance, helping you to keep a solid overview. ⚠️ Always Allow Mode If you're feeling dangerous and tired of always reviewing and approving shell tool calls, we now offer an (experimental and dangerous!) "Always Allow" Mode. 🎥 Remotion Plugin Our new Remotion Plugin makes it easy to create consistently styled videos with Remotion in stagewise.

Last week, we shipped six releases, focusing on much-requested features: ⚫️ GPT 5.5 support Once GPT 5.5 was available on the public API, we immediately shipped it as a new model in our app. OpenAI's latest frontier model comes with a new level of intelligence and agentic capabilities, but costs significantly more. If you need its new capabilities: The model is available in stagewise now! 👀 Agent Preview Panels Running multiple agents at once can quickly become a hassle. We shipped Agent Preview Panels to see an agent's work at a single glance, helping you to keep a solid overview. ⚠️ Always Allow Mode If you're feeling dangerous and tired of always reviewing and approving shell tool calls, we now offer an (experimental and dangerous!) "Always Allow" Mode. 🎥 Remotion Plugin Our new Remotion Plugin makes it easy to create consistently styled videos with Remotion in stagewise.

You wanted to know how we create our launch videos, so we created a guide for that! 🎥📕 We're showcasing how to get started using only stagewise (including it's new built-in Remotion plugin!), a small song for background music and a reference website that we turn into a launch video. Check it out on YouTube: https://lnkd.in/dui7xjxJ p.s. I know about the mic quality issues - won't happen again 😇

Last week, we shipped 9 releases for our AI coding agent. Here's everything you need to know: 💻 Persistent Shell Support stagewise can now start and manage proper shell sessions, meaning it can not just run single commands, but start and watch dev servers, build scripts and much more! 🟦 Kimi K2.6 Support stagewise now supports Moonshot.AI's latest and greatest model Kimi K2.6, offering almost SOTA performance at a fraction of the cost. 🔖 Renamable Agents You can now give your agents custom names by simply clicking on their title, making it much easier to always select the right agent. 🔥 Performance improvements We've identified and fixed multiple performance related issues, resulting in a much smoother experience especially for power users with significant amounts of chat data in stagewise. To read more, click here: https://lnkd.in/dze65xpq

Last week, we shipped 9 releases for our AI coding agent. Here's everything you need to know: 💻 Persistent Shell Support stagewise can now start and manage proper shell sessions, meaning it can not just run single commands, but start and watch dev servers, build scripts and much more! 🟦 Kimi K2.6 Support stagewise now supports Moonshot.AI's latest and greatest model Kimi K2.6, offering almost SOTA performance at a fraction of the cost. 🔖 Renamable Agents You can now give your agents custom names by simply clicking on their title, making it much easier to always select the right agent. 🔥 Performance improvements We've identified and fixed multiple performance related issues, resulting in a much smoother experience especially for power users with significant amounts of chat data in stagewise. To read more, click here: https://lnkd.in/dze65xpq

Introducing Kimi K2.6 support in stagewise Kimi K2.6 is the latest SOTA open source model from MoonshotAI, offering a very strong combination of: - Similar performance to Claude Opus 4.5 in selective areas - Cost level comparable to GPT 5.4 mini This unlocks a new level of cost-efficient agentic coding, and to celebrate this, we're giving away a limited number of 1 month free Pro subscriptions using the Coupon KIMI-K26-LAUNCH.

Introducing `/preview` in stagewise. Most design iteration happens in the actual codebase. stagewise now lets your agent spin up isolated preview apps on demand. One command: `/preview`. Try it free: https://lnkd.in/d9RZawSh #stagewise #devtools #webdevelopment #frontend #aiagent

Introducing First-Day support for Claude Opus 4.7 in stagewise 🚀 And to celebrate this occasion, we're offering 20x one month Pro Subscription for free using the Coupon OPUS-4-7-LAUNCH Just update to the latest alpha build to get started with Anthropic's latest high-end model. Pricing is the same as for Opus 4.6, which get's disabled by default but can be re-enabled in the settings if you wish to continue using the older model.

What if your coding agent could just learn? We just shipped the `/learn` command in stagewise. With `/learn`, you can turn docs, in-chat discussions, and corrections into a reusable skill with one command. That means less repeating yourself. Less re-explaining project context. Less losing good workflows between sessions. Link to the full announcement in the comments 👇