Zero-config toolchains, mise, and deploy.

appz

appz detects your stack (Node, Rust, Python, ...), wires up mise with cached, incremental tasks instead of a hand-rolled build cache, and drives install/build/dev/test/lint/format/deploy through one command — no config to write by hand.

• Zero-config detection — reads your project, not a config file, to know how to build it (96+ frameworks, mise tool versions, output dirs).
• mise-native, not another task runner — generates [tools] + [tasks] with sources/outputs, so mise's own cache — not a second one — decides what to skip.
• Unified deploy — appz deploy drives each platform's own CLI (Vercel, Netlify, ...) instead of competing with them.
• Built for agents too — appz mcp exposes detect/build/deploy as MCP tools, so an AI agent can drive the same pipeline a human does.

appz init      # detect toolchains, scaffold appz.jsonc + mise.toml
appz install   # mise + package manager install
appz build     # mise run build (skips when inputs are unchanged)
appz dev       # start the dev server
appz test      # run tests
appz lint      # run the linter
appz format    # run the formatter
appz deploy    # deploy via the target platform's own CLI
appz doctor    # diagnose project stack, config, and suggestions
appz mcp       # stdio MCP server exposing detect/doctor/run to AI agents

MCP server

appz mcp speaks the Model Context Protocol over stdio so AI agents can drive appz directly. Tools:

• detect — JSON toolchain report for a directory
• doctor — JSON diagnosis + suggestions
• run — execute a lifecycle command (install/build/test/lint/format) and return its output

Point an MCP client at appz mcp (e.g. {"command": "appz", "args": ["mcp"]}).

Layout

• crates/appz — the CLI binary
• crates/appz-core — toolchain detection, mise.toml generation, and the doctor/report logic appz runs on top of
• crates/appz-deploy — the DeployProvider trait + per-platform CLIs
• crates/command — shell-free process execution (arg-vector, timeouts)

Build

cargo build --release