Show HN: Memcode launches a new terminal coding agent

Memcode has entered public beta as a new developer platform, announced via a Show HN post. The product bundles a terminal coding agent, chat, reusable agents, DataHub, and a Lovable-style website generator into one stack aimed at day-to-day software work.

The coding agent is heavily inspired by Claude Code and is built to run across frontier models as well as open-weight options such as Kimi and GLM. Persistence is a core mechanic: the agent keeps memory and state recall so sessions stay continuous rather than resetting on every prompt. The broader platform also exposes chat, reusable agents, DataHub, and a website generator, so the agent sits inside a wider workflow surface instead of as a single-purpose CLI tool.

For engineers and builders, that combination targets the usual friction with coding agents that forget context, lock you to one model, or force you to re-establish project state each time. Model choice across frontier and open-weight backends matters for cost, latency, and data-control tradeoffs. Reusable agents and DataHub suggest shared workflows and structured project data rather than one-off chat threads.

The positioning is competitive with Claude Code–style terminal agents while adding multi-model support and coworker-style continuity. The Lovable-style website generator puts Memcode nearer product/prototyping tools as well as pure coding agents. The stated origin—frustration with existing agents—frames the bet as product and behavior, not just another model wrapper.

Watch how well persistent memory and state recall hold up on real multi-day codebases, whether reusable agents and DataHub become the sticky surface beyond the terminal agent, and how model routing across frontier and open-weight stacks (including Kimi and GLM) behaves under daily use. Public beta is the right window to stress those claims before the platform hardens its defaults.