Harper

An agent is a loop, and the loop is only as fast as the slowest service it depends on. Most stacks make the loop traverse the network. Vectors live in one service. Memory lives in another. Customer data lives in a third. Each turn is a round trip, and the round trips compound. Harper runs the loop in-process. Data, vectors, agent code, and the deterministic logic around the model share one runtime. Context assembles in under 50ms. The agent reads at memory speed instead of network speed. One runtime from local dev to global edge. No services to wire between prototype and production. See the architecture: https://lnkd.in/gnURFupf #AgenticEngineering #AIInfrastructure #LLMOps

Two wins. One runtime. 🏆 Open Source Data Platform of the Year 🏆 In-Memory Solution of the Year Out of thousands of nominations and dozens of categories, Harper is the only company to win in two categories. Open source earns trust. In-memory earns performance. The Harper's unified runtime delivers both in the same process. Alongside honorees like MariaDB, Aerospike, Yugabyte, Grafana, and VAST Data. Read the full announcement → https://lnkd.in/dD4RtACy

Geofencing is easy. Turning a proximity signal into a personalized offer is where many stacks fall apart: four or five services, latency at every hop, and an architecture that has to be hand-assembled piece by piece. Aleks built this entirely on Harper. Spatial lookup, customer data, campaigns, rules, and the Claude call, all from one runtime. One hop to the LLM. Zero hops for everything else. That’s the agentic runtime thesis in a surprisingly small amount of code.

Nobody wants to pick a data center. Bari Jay closes out Harper 5.0 launch week with the case for why deployment should be the last thing developers think about, not the first. The traditional cloud experience asks engineering teams to become infrastructure experts before they ship a single line of application code. Pick a provider. Pick a region. Pick an instance. Wire up the database, the cache, the messaging layer. Configure replication. Coordinate certificates. Hope it holds. Harper Fabric asks a different question. What does your workload look like? What regions are your users in? What performance do you need? You answer. Fabric does the rest. Five minutes to a live, globally distributed cluster. This is the piece that makes everything else we shipped this week usable at scale. Harper 5.0, our open source runtime, with a deployment experience built for scale without the headache. Read Bari's full breakdown: https://lnkd.in/g_jrRGwj #Harper5 #AgenticEngineering #CloudInfrastructure #DeveloperExperience

Shoutout to Chris Barber for shipping rocksdb-js. A ground-up Node.js binding for RocksDB, open sourced under Apache 2.0. The Node.js ecosystem has had some options for a while, but none that expose the full engine. Wrappers built on old LevelDB abstractions. Abandoned forks. Bindings without real transaction support or TypeScript. If you wanted to use RocksDB seriously from JavaScript, you were writing glue code or compromising on what the engine could actually do. rocksdb-js closes that gap. Full transaction support with optimistic and pessimistic isolation. Lazy range iteration. A native C++ binding on modern Node-API. Prebuilt binaries across macOS, Linux (glibc and musl), and Windows. A built-in transaction log system that powers replication in Harper Pro (The enterprise extension of Harper 5.0) and is ready for event sourcing, change tracking, or whatever the community builds next. Huge thanks to Chris for the craft and care he continues to pour into this library. Open source matters more now than ever. AI is compressing the distance between an idea and a working system, and the ecosystem moves fastest when the foundations are shared. Read the walkthrough: https://lnkd.in/gR4Zq4bR #NodeJS #OpenSource #RocksDB #DeveloperTools

Harper 5.0 is open source, because we believe the future of software, especially for AI agents, is a unified runtime, not a fragmented stack. In an AI-driven world, simplicity becomes a force multiplier. Today’s systems spread logic across databases, caches, APIs, and services. That fragmentation makes it harder for AI to reason about how things actually work. A unified runtime changes that. With fewer layers and a smaller surface area, AI has a clearer view of the system, and with access to the source code, it can reason more effectively, generate better outcomes, and move faster from idea to production. That’s a big part of why we made this move. We open sourced the core under Apache 2.0 so developers and AI can inspect it, extend it, and run it anywhere. Ethan Arrowood led the push behind this shift, and in the full breakdown, he walks through how we approached the dual license model between Harper and Harper Pro to balance open source with long-term sustainability. Read the full story ↓ https://lnkd.in/g5Yfn-kg #Harper5 #OpenSource #AgenticEngineering #SoftwareArchitecture

When your SVP of Engineering writes a 2,000-word technical deep dive on launch day, you read it. Kris Zyp designed Harper's Resource API to eliminate the translation layer between HTTP and application code. In Harper v5, that design gets sharper. Class methods map directly to HTTP verbs. The same interface handles requests and data access. No routing abstraction. No glue code bridging two systems. For engineering teams, that means fewer layers to debug, fewer abstractions to onboard, and fewer things to break. For the business, that is faster delivery and a smaller maintenance footprint. The source resource caching model is where it gets really interesting. Worth the read if your team is managing any kind of tiered data architecture. Stay tuned for more. This is only day two of Harper 5.0 launch week. https://lnkd.in/gYGPX-ME #Harper5 #AgenticEngineering #SoftwareArchitecture #OpenSource

We sat down with Bari, Kris, and Ethan to talk through everything behind Harper 5.0. The open-source decision. The RocksDB engineering effort. Why a unified runtime changes how agents go from prototype to production. No changelogs here. Just three people who built the thing, explaining why it matters. Watch the video. Read the full breakdown. Link below. This is also day one of a full week of Harper 5.0 content. We have deep dives coming on the open-source story, the agentic engineering thesis, RocksDB internals, and more. If today's release caught your attention, stick around, follow Harper. There is a lot more to unpack. https://lnkd.in/ghspuQwy #OpenSource #AgenticEngineering #Harper5 #SoftwareArchitecture

Harper 5.0 is live. This is the release we have been building toward. Harper's core runtime platform is now fully open source under Apache 2.0. Harper Pro is source available under ELv2. RocksDB ships as a native storage engine alongside LMDB. Join some of the world's largest companies and start building with Harper today. Harper is designed to collapse database, cache, messaging, and application logic into one runtime, delivering extreme performance, cost-efficiency, and scale. Now the entire JS community can see how it works, build on it, extend it, or contribute back. Read the full announcement below. https://lnkd.in/ghYQdbzZ

New Podcast Episode: The Future of Nuxt & Full-Stack Development with Daniel Roe We’re excited to share our latest podcast episode featuring Daniel Roe, a full-time open-source contributor and lead of the Nuxt framework. In this conversation, Daniel dives into: - The evolution from Nuxt 3 → 4 → the upcoming Nuxt 5 - What it really takes to maintain and scale a major open-source framework - Upgrade strategies for production apps - The future of Vue-powered full-stack development If you're building modern web applications, leading engineering teams, or working in the Vue/Nuxt ecosystem, this is a discussion you won’t want to miss. Daniel offers candid insights into versioning, long-term maintenance, and what’s next for one of the most important frameworks in the Vue community. Watch the full episode https://lnkd.in/gHTV_z2H #Harper #Nuxt #VueJS #WebDevelopment #OpenSource #FullStack #JavaScript #TechLeadership