Use in your project
Install the AI agent sandbox dev container template into any project with the Convoy CLI.
The Convoy dev container is also available as a standalone template you can install into any project — it does not require Convoy. You get Claude Code, Codex CLI, OpenCode CLI, and Cursor CLI preinstalled in an isolated container, with an optional outbound-network firewall (strict / open modes).
Install with the CLI
From your project root:
npx @useconvoy/cli devcontainerOr, if you already have the CLI installed:
convoy devcontainerThis writes four files into ./.devcontainer/:
devcontainer.jsonDockerfileinit-firewall.shREADME.md
The command fetches the latest template from GitHub when reachable, and otherwise falls back to the copy bundled with the CLI package. Pass a directory argument to target another project (convoy devcontainer ../other-project) and --force to overwrite an existing .devcontainer/.
If a Convoy CLI profile resolves for the project, the command also prints the exact firewall value to export for strict network mode — CONVEX_DEPLOYMENT=<slug> for Convex cloud backends, or DEVCONTAINER_EXTRA_ALLOWED_DOMAINS=<host> for self-hosted ones — so you don't have to derive it from your backend URL.
Install manually
Copy the files from templates/devcontainer/ in the Convoy repository into a .devcontainer/ folder at your project root.
What's Convoy-specific
Nothing is required. Two things light up automatically for Convoy users:
~/.config/convoyis mounted into the container, so CLI profiles and API keys from the host work inside it- setting
CONVEX_DEPLOYMENTon the host adds your Convex backend to the strict-mode firewall allowlist
Everything else — agent installs, shared Codex/OpenCode auth mounts, network modes, DEVCONTAINER_EXTRA_ALLOWED_DOMAINS for your own API endpoints — is generic. The generated .devcontainer/README.md documents all environment variables and modes.