An agentic coding CLI built for long sessions. Three things make it different: Chitta - persistent memory, so it actually remembers your codebase between sessions instead of starting cold. UltraGuard - built-in prompt-injection and PII shield. Three-tier context compaction - it summarises its own history mid-task, so it doesn't lose the plot on big refactors. GLM 5.2 is unlimited right now. npm i -g 100xprompt-cli

Hey Product Hunt! I'm Nipurn, maker of 100xPrompt CLI. I got tired of two things with coding agents: they forget everything between sessions, and they fall apart halfway through long tasks. So we built ours around memory. Chitta gives the agent persistent memory of your codebase - conventions, past decisions, structure - so day 2 doesn't start from zero. Three-tier context compaction means it summarises its own history mid-task instead of silently dropping context on big refactors. And UltraGuard sits in the request path blocking prompt injections and masking PII before anything leaves your terminal. We run our own inference stack (speculative decoding, prefix caching, KV compression), which is how we can do this: GLM 5.2 is UNLIMITED right now - no caps, no queue. npm i -g 100xprompt-cli Break it and tell me how. I'm here all day.

The three-tier context compaction for big refactors is the interesting part, does the CLI let you point it at a custom Anthropic-compatible or OpenAI endpoint, or is it locked to GLM 5.2? Curious how Chitta's persistent memory holds up when swapping models mid-project.

100xprompt

Agentic coding CLI with unlimited GLM 5.2. Free right now

Agentic coding CLI with unlimited GLM 5.2. Free right now