
Devspace
Your smartest, fastest teammate. They’re called Astroagents.
47 followers
Your smartest, fastest teammate. They’re called Astroagents.
47 followers
Astroagents work like a full-stack product team inside your browser. Tell them the idea and they handle specs, code, UI, backend, and deployment.







Hi everyone 👋
Thanks for taking a look at Devspace.
We built Devspace because most AI building still feels like guessing. You describe an idea, the AI misinterprets it, and you keep refining prompts hoping it gets closer. It’s frustrating, and it slows down creativity.
So we took a different approach.
We’re building a team of AI Astroagents. Teammates that work with you to understand, design, and build your webapps.
But for the beta, we’re starting with just one:
Meet Sol
Sol is our all-rounder agent the one who:
Understands your idea from natural language
Plans the UI and flow
Generates clean, editable code
And lets you refine everything directly in your browser
No setup. No terminals. No local environment.
Just start building.
The rest of the Astroagent team will be rolling out gradually, based on what works best for you and what you need most.
We want to develop them with your feedback, not just drop them all at once.
And for the beta, we’re going all in:
It’s completely free. Unlimited usage.
No paywalls. No limits.
No API key. No card. Just create and build.
We’d love for you to try it, experiment, and tell us where it feels magical vs where it still needs work.
Your feedback is literally what shapes the next stage.
Thanks for being here 🙏
Himayun, Founder of Devspace
Cal ID
Congrats on the launch!
How do the agents coordinate with each other when working on different parts of the same project?
@sanskarix Thanks so much, Sanskar! 🙌 Appreciate you asking.
Right now we have Sol as the all-rounder agent. she handles the entire workflow herself.
But we’re rolling out a team of specialized agents soon:
Product Manager Agent – understands what you're trying to build
AI Engineer Agent – handles the architecture + model integration (especially useful when users want to build AI-powered products)
QA Engineer Agent – tests, catches issues, and suggests improvements
They all work from a shared project spec, so everyone is aligned.
And the cool part:
When one agent finishes their step, they literally “hand off” the context to the next agent including decisions, reasoning, and changes made.
So they talk to each other about workflows and implementation details, ensuring no one overwrites or conflicts with others.
Think of it like a mini product team where each member says, “Okay, my part’s done — your turn.”
Super excited to share more once the full agent squad rolls out!
Cal ID
@himayun That's really smart, and it saves a lot of potential clashes. Also, the QA engine agent sounds interesting.
All the best for the launch and keep building things up!
@sanskarix Thanks a ton, Sanskar! 🙌
Yeah exactly avoiding clashes was a big design goal. The QA agent is going to be super helpful especially when projects start getting complex. it’ll catch weird edge cases and suggest fixes before the user even notices.
Really appreciate the support!
Swytchcode
Congratulations on the launch!
Do you also deploy the code or is it code generation only for the time being?
@chilarai Thank you! 😊
Yes. it’s not just code generation. Devspace handles the full flow: build → iterate → preview → deploy, all in one place.
You can take an idea to a live app without touching configs, API keys, or deployment setups. And for now, everything is completely free. we’re covering all infra during beta.
Happy to hear your thoughts if you try it out!
Congrats on the launch!
I really like how natural the prompting feels. Do you have any recommended prompt examples for more complex apps?
@davidoneby Thank you! We’ve tried really hard to make the prompting feel like you’re just explaining your idea to a teammate.
For more complex apps, we usually guide users to start with the goal + the core actions rather than trying to describe UI details. For example:
Instead of:
“Build me a job board platform with multiple pages and filters…”
Try:
“I want a job board where recruiters can post jobs and candidates can apply.
Main actions: create job, edit job, apply, view applicant list.
Let’s start with a clean 2-page layout.”
From there, the agents will handle:
Structure
Components
Data models
API routes
And you can refine step-by-step in natural language.
We’re also adding a Prompt Library inside Devspace soon real examples built by users
Love the concept of Astroagents as teammates! For teams building user-facing apps, how do these agents ensure the code they generate is accessible?
Things like proper semantic HTML, ARIA attributes, keyboard navigation often get missed by AI.
Great launch!
@halyticai Thank you! That’s a really good point accessibility is something we take seriously.
Our QA Agent is responsible for reviewing the generated UI and making sure it follows accessible patterns (semantic HTML, ARIA roles, correct labels, sensible tab order, etc.). When the UI is generated, the QA Agent actually reviews and communicates back to the Implementation Agent if something is missing. So there’s a built-in feedback loop rather than just one-shot code generation.
We’re also adding:
Automated accessibility checks (Lighthouse + axe-core)
Design system defaults that are accessible by construction
This way, teams don’t have to manually enforce accessibility every time it becomes the default.
Really appreciate you calling this out.
We want to make accessibility a first-class part of dev, not an afterthought.