◆
Portable memory
Import your memory from Claude Code & Codex. Claims, episodes and procedures travel with you — the agent remembers across sessions and machines.
███████╗ █████╗ ███╗ ██╗ █████╗ ╚══███╔╝██╔══██╗████╗ ██║██╔══██╗ ███╔╝ ███████║██╔██╗ ██║███████║ ███╔╝ ██╔══██║██║╚██╗██║██╔══██║ ███████╗██║ ██║██║ ╚████║██║ ██║ ╚══════╝╚═╝ ╚═╝╚═╝ ╚═══╝╚═╝ ╚═╝
Memory you can import. Agents that run while you sleep. Typed tools
instead of blind shell, and real permissions on every call. A single
Bun monorepo — chat TUI, headless zana run,
and a background daemon.
$ curl -fsSL https://zanaagent.app/install | sh macOS & Linux · self-updating · MIT licensed
Five things the generalists don't give you.
Import your memory from Claude Code & Codex. Claims, episodes and procedures travel with you — the agent remembers across sessions and machines.
A real daemon runs scheduled and autonomous agents. Close the loop overnight: scan → fix → open a PR while you sleep.
Many machines, one event bus. Your laptop and your VPS run as one agent over TLS — delegate work across the mesh.
Toolpacks with typed schemas and declared side-effects instead of blind shell calls. The LLM is the primary user of its tools.
Capability gating on every single tool call. Read-only means read-only — the security model is enforced where the tool runs.
Runs on your own providers — Anthropic, OpenAI, Codex, Copilot, DeepSeek, Ollama, OpenRouter. No mandatory cloud. Your keys, your machine, your data.
The TUI, delegating agents in real time.