codeflower interactive

Start a sandbox for a project with one agent, to work with from VS Code. Each folder is mounted in a subfolder named after it: data folders read-only in /data/input, output folders read-write in /data/output, code folders (your repositories and packages, which the agent may change) read-write in /data/code, and skills folders read-only in /data/skills, where the agent loads them from.

With –isolate-network the sandbox has no network of its own: the app is its only way out, and lets it reach only its agent’s service and the hosts in the allowlist file. Edit the file while the sandbox runs to allow more.

It prints a link to open in the browser while “codeflower client” runs on your computer, which adds the sandbox there; or, in VS Code, run “code flower: Connect to worker” and paste the codeflower://… line. If the agent is not signed in for this project yet, the page or VS Code offers its sign-in after connecting. Press Ctrl+C to stop.

With –server the sandbox also serves its chat page on the same port and prints a link to it (http://HOST:PORT/?key=…), like Jupyter’s: open it in any browser that reaches this machine, e.g. through ssh -L. Anyone with the link controls the agent, and plain HTTP carries the key and the chats unencrypted, so keep job logs private and prefer –host 127.0.0.1 with a tunnel on shared networks.

To start a session with an initial task and continue it in VS Code, see “codeflower run”. To let the agent work on a task by itself instead, see “codeflower task”.

Usage

codeflower interactive [flags]

Examples

  codeflower interactive --agent claude --input ~/data/study42 --output ~/results/study42
  codeflower interactive --agent claude --input ~/data/rna --input ~/data/atac --code ~/src/mypkg
  codeflower interactive --agent codex --project study42 --input /data/study42
  codeflower interactive --agent claude --input ~/data/study42 --skills ~/my-skills --isolate-network
  codeflower interactive --agent claude --image ./codeflower-sandbox.sif --state ~/cf/study42 --bind /data/ref:/reference

Flags

Flag	Default	Description
--advertise-host string		address printed for VS Code and the page link (default: this node’s IPv4 address; 127.0.0.1 with Docker or Podman)
--agent string		agent the sandbox runs: claude, codex, dsh, or opencode (or set $CODEFLOWER_AGENT; asked if left out)
--allowlist string		file of hosts an isolated sandbox may reach, one per line (default: PROJECT.allowed-hosts.txt next to the project)
--bind stringArray		more mounts: SOURCE[:TARGET[:ro|rw]] (repeatable; default target /data/input/NAME, read-only)
-c, --code stringArray		repository or package the agent develops, mounted read-write at /data/code/NAME (repeatable)
--config string		config directory with config.yaml and skills (or set $CODEFLOWER_CONFIG; default ~/.codeflower)
--cpus int		CPUs the agent may use (default: the batch job’s allocation; Docker and Podman also enforce it)
--engine string	auto	container engine: auto, docker, podman, or apptainer (or set $CODEFLOWER_ENGINE)
--host string	0.0.0.0	interface the worker listens on
--image string		sandbox image: a Docker/Podman image name, or an Apptainer .sif file (or set $CODEFLOWER_IMAGE; default ghcr.io/lutrarutra/codeflower-sandbox:0.15.1, downloaded on first use)
--image-tag string		build of the published image to use instead of 0.15.1, the release this app is made for: latest, another release (0.2.1, or 0.2 for its newest patch) or a commit (sha-abc1234) (or set $CODEFLOWER_IMAGE_TAG)
-i, --input stringArray		folder with your data, mounted read-only at /data/input/NAME (repeatable)
--isolate-network		cut the sandbox off the network except for its agent’s service and the hosts in the allowlist (Linux)
--memory string		memory the agent may use, e.g. 32G (default: the batch job’s allocation; Docker and Podman also enforce it)
--no-tunnel		keep this sandbox from browsers paired through the tunnel (it does not start the client for the tunnel either)
-o, --output stringArray		folder for results, created if needed and mounted read-write at /data/output/NAME (repeatable)
--port int		port of the worker (and its page with –server) (default: any free port)
-p, --project string	default	project name: its sessions, logins and workspace live in ~/codeflower/projects/NAME
--server		also serve the chat page on the worker’s port, and print a link to open it in a browser
-s, --skills stringArray		folder with a skill (a SKILL.md) or with skill folders, mounted read-only at /data/skills/NAME (repeatable)
--state string		project directory, instead of –project (created with owner-only permissions)
--time-limit string		time the agent has, e.g. 8h or 1-12:00:00 (default: the batch job’s time limit); reported, not enforced