@GetAgentHail · X

AgentHail — a control layer for AI agents to request approval, execute work, and return verifiable results. Would you sign up or leave?

interface1860 · HN

TaskFuel – agents discover and pay per call for 100 APIs

@bonduelleioat · X

YOU CAN STOP PAYING $30–100+ EVERY MONTH FOR AI VIDEO GENERATION. Expensive subscriptions like Higgsfield can be replaced with a local agent built on Claude Code + the generation aggregator, allowing you to create viral faceless vertical videos for around $0.75 per video - with no monthly subscription and no credits that expire. The system follows a simple principle: you only pay for the result. A specialized /generate skill automatically selects the cheapest model for generating the base asset, tracks spending through an internal ledger, checks aspect ratios before animation, and sends the finished content to Blotato via MCP for multi-platform distribution. Instead of paying for an expensive AI subscription, you essentially get your own mini content factory that creates videos whenever you need them. And the most interesting part: $0 subscription. ~$0.75 per finished video. Zero wasted credits. This is how AI is moving from “pay for access” to “pay only for the

@RunPei83973 · X

Claude Code users: Did you know API routing can significantly reduce AI costs? We built VIAPI to test different AI routes: Budget → lower cost Stable → production Pro → high availability

g418572664 · V2EX

做了一个文档解析与记忆工具，专门辅助给传统行业做 AI 落地的老哥 现在 AI+的工作还挺常见的，就像大佬们说的，“所有行业的产品可能都会用 AI 重新做一遍”。最常见的就是各种 agent ，说要用 AI 赋能传统行业啥的，代替人类专家去处理海量的复杂资料、进行深度分析并做出决策。 举个例子，金融行业的“智能审计与尽调 Agent”。 过去，银行或投资机构想要给一家企业贷款或投资，需要人类审计师去读几十份、每份几百页的招股书和财务报表。现在虽然有了 AI ，但把文件一股脑全丢给它是不现实的，且不说烧 token 的问题，这些文档里有无数的跨行、跨列单元格表格，普通工具一拉，表格数据全串行了。如果 AI 把“第一季度利润”和“第二季度支出”的信息碎在一块，那得出的财务分析就完蛋了。 所以，现在要真想开发出一个能干活，还确保正确率的 agent ，就需要一个专业的、AI-native 的解析工具，把复杂的表结构和章节层级完整还原出来。我做的工具 Knowhere 就是干这个的： https://knowhereto.ai/?utm_source=v2ex 它能把复

osaitech · Product Hunt

OpenSourceAI.tech Discover 10,000+ open-source AI projects, models & tools

bingus-bongo · HN

Use GLM-5.3 in Cursor today via tokengo API

@trycaedral · X

@BobbySBX · X

We just dropped Assay v0.2.6, the open source evaluation engine powering Most registries judge an AI tool by its README. We built Assay to actually run skills, MCP servers, and plugins inside isolated sandboxes with live models to observe what they do before you install them. What is new in v0.2.6: - NVIDIA API Catalog provider: Native support is here, allowing you to use NVIDIA driver and judge models to run behavioral benchmarks and multi-tier audits. - Zero runtime dependencies: Still a self-contained ~240 KB tarball that speaks directly to @nvidia , @AnthropicAI , @OpenAI , @OpenRouter , and local HTTP endpoints without SDK bloat. - Adversarial and static checks: Out of the box verification against prompt injection, tool scope creep, destructive shell commands, and supply chain drift. Every tool on is tested through Assay. That means 4,200+ skills, MCP servers, and plugins are commit-pinned, graded on actual runtime behavior, and rea

tahazsh · HN

Hi! I’m Taha. In many agentic products that support workflows (including one I worked on), I noticed they either don’t support node-based editors, or use React Flow and go through the difficult work of integrating it into their product to run it and work with their existing logic. So I thought about creating a tool that could help with this by closing the gap between the editor and the runtime. That’s why I created Wayflow. The basic architecture is simple: you just need to create a graph (which is a JSON object) that the runtime knows how to run. The runtime doesn’t care where that graph is coming from, it just needs the right schema. And with the help of the editor, you can create the graph, and then export it or directly save it on your backend in your database. And then when you want to execute it, you just hand it to the runtime. The runtime can either stream the execution (which is useful for the editor), or give you the final result. How you execute the graph is up to you: t