Intent - Describe a feature and AI agents build, verify, and ship it
by•
Intent is a developer workspace built for agent-driven development. Define a feature as a spec, and a team of agents coordinates the work (from implementation to verification) inside an isolated workspace with built-in code, terminal, and git.
Replies
UXPin Merge
Really interesting direction. Moving from copilots to coordinated agents working from a spec feels like a big shift. How do you see teams defining good specs so the output stays reliable?
@uxpinjack The shift only works if you treat the spec as a living contract between humans and agents, not a fancy prompt.
What we’re seeing work in Intent:
Write for agents: structure, constraints, edge cases, and verification steps, not tutorials or giant summaries.
Keep it living: agents read from and write back to the spec as they work, and a verifier agent checks output against that spec before anything ships.
When teams treat spec quality and spec review as seriously as code review, reliability goes way up.
Honestly
Congrats on. the launch! What is the feasibility of adjusting the number of & type of agents you want for a given project, or is everything primarily decided for you?
@scott_davidson_jr Everything is in your hands. We’ve set up some personas for you right out of the box, but you’re free to modify them and add new ones. Intent will read, understand, and use them.
@jaysym I see that it's free if we use our Claude-code/codex. How's the pricing decided if we enable the context engine for non-Auggie agents?
@sandeepan_jindal This is correct: you can use Intent with your Claude Code or Codex subscription without paying anything extra or having an Augment Code account. If you ever decide to activate the context engine with one of these providers, that’s when you’ll need to create an Augment Code account. There is somes free context engine call you can make, and after that you’ll just pay the small amount of credits required per call.
The context engine can help you save tokens with other providers while improving output quality, so in my opinion the cost is absolutely worth it. You can experiment with it on and off to see how many credits it uses and then decide later. If you have more questions, I’ll be happy to help.
@jaysym Thank you. That's what I thought, but there was no transparency in the pricing. I enabled it with Claude Code yesterday and don't see any reduction in the auggie credits. But as you said, there'll be a cost in the future, so no way to know right now.
The isolated workspace piece is interesting. If the agents make a mess mid-build (bad merge, broken dependencies, half-written migrations), what's the recovery story? Does the isolation mean you can just nuke and restart, or are you expected to debug whatever state they left behind?