Zhipu launches ZCode 3.0, an agentic coding harness tuned for GLM-5.2

ZCode is z.ai’s agentic coding environment, now at version 3.0 and optimized for the GLM-5.2 model. It positions itself as a full plan-code-review-deploy loop that plugs into 20+ existing coding tools, adds multi-agent collaboration, and can be triggered and steered from chat apps like WeChat, Feishu, and Telegram. The pitch is squarely aimed at the current ‘vibe coding’ market, with a demo of building a Gomoku game as the hero example.

The offering is sold through tiered GLM Coding Plans rather than per-token pricing: a Lite tier at $16.20/month for small repos, a Pro tier at $64.80/month with 5x the usage and curated MCP tools, and a high-volume tier at $144/month with 20x usage and priority model access. All tiers advertise rolling access to z.ai’s latest flagship models.

Worth noting for readers: this is a product landing page, not independent reporting or benchmarks. The claims about GLM-5.2 reasoning, multi-agent performance, and ‘faster, steadier’ agentic coding are the vendor’s own, and there’s no third-party evaluation here to verify how ZCode stacks up against established harnesses like Claude Code, Cursor, or Codex.