面向使用 AI 编码智能体的开发团队，自动为无文档代码库生成可信规格说明，让规格审查成为新的合并请求审查，减少人工看护成本。
热门评论
PH 用户
Hey Product Hunt 👋
I'm Xesca, CTO turned founder. Before EasySpecs, I spent 15+ years being the person accountable for code I didn't personally write, across fintech, insurtech, and media.
That's the job now for every engineering leader: your agents ship code faster than anyone can read it, and you're still the one who answers for what breaks.
The problemI kept hitting the same wall everywhere I looked:Teams adopt AI coding agents, and PRs pile up faster than anyone can review themSpecs get written against a mental model of the codebase, not the actual codebaseAgents generate code against those specsThings break silently, and nobody can trace why, because nobody documented what was actually there to begin withCode review is dead. Spec review is where quality lives now. But spec review only works if the spec is grounded in reality, not guesswork.What EasySpecs doesTwo things, one workflow:
1. We document your codebase automatically
Even the undocumented, nobody-wants-to-touch-it kind. Our first real test was a pharmaceutical company's legacy Delphi system. If it works there, it works on whatever you're building.
2. We turn that understanding into specs your team and your agents can actually trust
Every spec comes paired with validators, so you know it matches the app before anyone codes against it.
Here's the part I'm most proud of: as you write a spec, the specs crafter suggests oracles and rubrics for it, concrete, checkable ways to verify the change actually did what the spec said.Oracles — the "does this pass or fail" checks a machine can runRubrics — the judgment calls that need a human or an AI reviewer to score against clear criteriaYou're not staring at a blank page trying to guess how to verify your own spec. The tool proposes it alongside the spec itself.Built for more than developersYou don't need to be a developer to use it. Right now the only way to sanity-check a spec against reality is to go read the source yourself.
We built EasySpecs so product managers and team leads get that confidence without opening a single file.Try itFree to start — connect a repo, bring your own AI key, see what it findsWorkbench — €5/repo/month while launch pricing lasts (normally €10)Factory — €150/month if you want us to run the AI side too, with credits to use on any repoI'd genuinely love your feedback, especially if you throw a messy, undocumented codebase at it. That's the case we built for.
PH 用户
FINALLY!!!! 🍾🎉🎊
PH 用户
Didn't know that these sort of tools existed. Excited to see this in action going forward
PH 用户
reviewing specs instead of 10k lines of agent code makes alot of sense 👍 are teams adopting it as the actual merge gate?
PH 用户
Congratulations @xesca_i_am @carlos_guirao & team @easyspecs.ai 🚀. Question, Having written the specifications, what would be the way to keep them consistent with the changing source code? Manually running them, or keeping them constantly up-to-date?
I'm Xesca, CTO turned founder. Before EasySpecs, I spent 15+ years being the person accountable for code I didn't personally write, across fintech, insurtech, and media.
That's the job now for every engineering leader: your agents ship code faster than anyone can read it, and you're still the one who answers for what breaks.
The problemI kept hitting the same wall everywhere I looked:Teams adopt AI coding agents, and PRs pile up faster than anyone can review themSpecs get written against a mental model of the codebase, not the actual codebaseAgents generate code against those specsThings break silently, and nobody can trace why, because nobody documented what was actually there to begin withCode review is dead. Spec review is where quality lives now. But spec review only works if the spec is grounded in reality, not guesswork.What EasySpecs doesTwo things, one workflow:
1. We document your codebase automatically
Even the undocumented, nobody-wants-to-touch-it kind. Our first real test was a pharmaceutical company's legacy Delphi system. If it works there, it works on whatever you're building.
2. We turn that understanding into specs your team and your agents can actually trust
Every spec comes paired with validators, so you know it matches the app before anyone codes against it.
Here's the part I'm most proud of: as you write a spec, the specs crafter suggests oracles and rubrics for it, concrete, checkable ways to verify the change actually did what the spec said.Oracles — the "does this pass or fail" checks a machine can runRubrics — the judgment calls that need a human or an AI reviewer to score against clear criteriaYou're not staring at a blank page trying to guess how to verify your own spec. The tool proposes it alongside the spec itself.Built for more than developersYou don't need to be a developer to use it. Right now the only way to sanity-check a spec against reality is to go read the source yourself.
We built EasySpecs so product managers and team leads get that confidence without opening a single file.Try itFree to start — connect a repo, bring your own AI key, see what it findsWorkbench — €5/repo/month while launch pricing lasts (normally €10)Factory — €150/month if you want us to run the AI side too, with credits to use on any repoI'd genuinely love your feedback, especially if you throw a messy, undocumented codebase at it. That's the case we built for.