Yitao Zhang

OpenClacky is looking to meet thoughtful investors in AI agent infrastructure, developer tools, open source, and enterprise workflow automation. Thesis: the first agent-runtime wave proved agents can act. The next wave is about cost-per-completed-task. As agents evolve from short-task tools into persistent workers, token efficiency becomes infrastructure, not a discount. OpenClacky is building a cost-efficient open-source agent runtime and local AI workbench: Web UI, CLI, memory, Skills, scheduled tasks, IM integration, browser automation, cache-aware execution, and BYOK/model routing. The wedge is an open-source runtime that can run agentic work cheaply, predictably, and repeatedly. Public product/GitHub signals: MIT open source, 946 GitHub stars, 16 core tools, BYOK/model freedom, Skill self-evolution, IM integration, desktop installer, and a public benchmark narrative around lower token cost than peer agent runtimes. The site lists existing backing from MiraclePlus, ZhenFund, Yingdong Capital, Variable Capital, Sequoia China, and Hillhouse Ventures. Proof points from our materials: 2,286 MAU, 1,082 WAU, 401 DAU; 367,562 agent tasks in the last 30 days and 111,495 tasks in the last 7 days; high task intensity around 42 tasks per DAU per day; production workload with 94.7% cache hit rate, 122,260 cached requests, and 4.43B prompt tokens processed in a single day. Public benchmark positioning: OpenClacky $5.10 / 51 req / 90.6% hit vs Claude Code $5.49 / 70 / 95.2%, OpenClaw $15.70 / 81 / 88.7%, Hermes $30.14 / 218 / 60.3% across three comparable tasks. Business model: C-side adoption through open source and power users; AI Keys/model routing for stable access, smart caching, unified billing, and usage growth; enterprise private agents, encrypted Skills, per-seat licensing, and workflow assessment/deployment services; long-term Skill creator economy. Public links: https://www.openclacky.com and https://github.com/clacky-ai/openclacky Warm intro line: I’m building OpenClacky, an open-source agent runtime focused on cost-efficient execution for long-running agent workflows. We think the next bottleneck is cost-per-completed-task, and we’d like to meet investors who understand AI infrastructure, devtools, and open-source distribution.