Start with the mental model
repository → basecamp sync → ready sandbox → engineer / agent → native tools
Basecamp is an environment bootstrap layer, not a sandbox and not a build system. It prepares the runtime and project environment before the useful workload starts.
Read the sandbox integration model →
Start here
- Install Basecamp.
- Run
basecamp initonce in an existing project. - Commit the generated
basecamp.lockwith the project. - In another machine, CI job, or sandbox setup phase, run
basecamp sync. - Use
basecamp shellwhen you want an interactive activated shell. - Run the project's normal build/test commands or hand the ready environment to a coding agent.
Core commands
| Command | Purpose |
|---|---|
basecamp init | Detect the project, clarify missing choices, create the first lock, and sync it. |
basecamp sync | Reproduce the existing lock without changing it. |
basecamp update | Re-resolve the project, update the lock, and sync the result. |
basecamp shell | Enter the synchronized environment. |
basecamp status | Inspect lock, project inputs, and materialized layers. |
basecamp doctor | Diagnose environment problems. |
basecamp clean | Remove project-local state; --cache also removes the shared store. |
basecamp bundle pack / unpack | Move a completely offline materialized environment. |
Polyglot model
The architecture is language-agnostic. Rust / Cargo is the complete v0.5 implementation; Node / pnpm, Python / uv, Go, and other stack adapters follow the same environment lifecycle.
AI environments
The environment graph is designed to grow beyond language dependencies into AI-facing inputs such as model assets, MCP servers, browsers, native/GPU SDKs, and datasets. These roadmap capabilities are kept distinct from what v0.5 already ships.