ikctl 1.4.2

CLI tool for deploying software to remote and local servers via SSH/SFTP.

ikctl — Install Kit Control

CLI tool to deploy and execute bash scripts ("kits") on remote servers via SSH/SFTP or on the local machine. Supports parallel execution, DAG-based pipeline orchestration, and git-based kit repositories.

How it works

A kit is a folder containing bash scripts and an ikctl.yaml manifest. ikctl uploads the scripts to the remote server via SFTP and executes them over SSH.

A pipeline is a YAML file that chains multiple kits across different server groups, with dependencies, parallel execution and output passing between steps.

Everything is driven by ~/.ikctl/config, which defines one or more contexts (server groups, kits source, secrets).

~/.ikctl/config          ← main config: contexts, active context
~/kits/
  config.yaml            ← server groups
  show-date/
    ikctl.yaml           ← kit manifest (uploads + pipeline + outputs)
    date.sh              ← the actual script
~/kits/pipelines/
  deploy.yaml            ← pipeline YAML

Kits are auto-discovered — no index file needed. Any folder under path_kits that contains an ikctl.yaml is a kit.

Requirements

• Python 3.13+
• git (for git-based kit repositories)

Installation

Recommended — pipx (isolated, no system conflicts):

pipx install ikctl

With uv:

uv tool install ikctl

With pip (on Debian/Ubuntu you may need --break-system-packages):

pip install ikctl --break-system-packages

From source:

git clone <repo>
cd ikctl
uv sync
uv run ikctl --version

Initial setup

On first run, ikctl creates ~/.ikctl/ and ~/kits/ automatically.

Minimal ~/.ikctl/config:

context: production
contexts:
  production:
    path_kits: ~/kits               # where your kits live (auto-discovered)
    path_servers: ~/kits            # where config.yaml lives
    path_secrets: ~/.passwords/pw   # file containing the sudo/ssh password
    path_pipelines: ~/kits/pipelines # optional: pipelines directory
    mode: remote
    timeout_connect: 30.0           # SSH connect timeout (seconds)
    timeout_exec: 120.0             # local command timeout (seconds)
    exclude:                        # optional: kits to hide from discovery
      - internal/debug-kit

Git-based kits (optional)

Pull kits from a git repository instead of a local directory:

contexts:
  production:
    kits_repo: [e-mail]:company/kits.git   # SSH repo
    kits_ref: main                                # branch, tag or commit
    # kits_token: $GITLAB_TOKEN                  # for private HTTPS repos
    path_servers: ~/kits/production
    path_secrets: ~/.passwords/password
    mode: remote

For private HTTPS repos:

    kits_repo: https://gitlab.com/company/kits.git
    kits_token: $GITLAB_TOKEN    # references an env variable

export GITLAB_TOKEN=glpat-xxxxxxxxxxxx
ikctl --list kits   # clones on first use, pulls on subsequent runs

Server config

# ~/kits/config.yaml
servers:
  - name: workers
    user: ubuntu
    hosts: [10.0.0.10, 10.0.0.11, 10.0.0.12]
    port: 22
    pkey: ~/.ssh/id_ed25519        # key-based auth

  - name: master
    user: ubuntu
    hosts: [10.0.0.5]
    port: 22
    pkey: ~/.ssh/id_ed25519

  - name: db
    user: root
    hosts: [10.0.0.20]
    port: 22
    password: $PASSWORD             # password auth (env var)

SSH authentication

Config key	Auth method	Notes
pkey: ~/.ssh/id_ed25519	Public key only	Ed25519, ECDSA or RSA
password: $PASSWORD	Password only
Neither	SSH agent + ~/.ssh/ discovery

pkey and password are mutually exclusive — only one is attempted.

Kit structure

# ~/kits/kubeadm-init/ikctl.yaml
kits:
  uploads:
    - kubeadm-init.sh     # files to upload to the remote server
  pipeline:
    - kubeadm-init.sh     # scripts to execute in order
  outputs:                # optional: documents KEY=VALUE the script emits
    JOIN_TOKEN: "kubeadm join token for worker nodes"
    JOIN_ENDPOINT: "API server endpoint (IP:6443)"

The outputs: field documents what KEY=VALUE lines the script writes to stdout. These values are available to subsequent pipeline steps via {{ steps.<id>.KEY }}.

Usage

ikctl [-l {kits,servers,context,mode,pipelines}] [-i INSTALL] [-n NAME]
      [-p [PARAM ...]] [-s {sudo}] [-c CONTEXT] [-m {local,remote}]
      [-v] [--dry-run] [--debug] [--describe KIT]
      [--pipeline FILE] [--timeout-connect F] [--timeout-exec F]
      [--parallel-workers N]

List

ikctl --list kits       # all discovered kits (shows declared outputs)
ikctl --list servers    # configured server groups
ikctl --list context    # active context and all paths
ikctl --list pipelines  # pipeline YAML files in path_pipelines

Inspect a kit

ikctl --describe kubeadm-init

Shows the kit's scripts and declared outputs:

╭─ Kit ───────────────╮
│  kubeadm-init       │
╰─────────────────────╯
Uploads:  kubeadm-init.sh
Pipeline: kubeadm-init.sh
Outputs:
  JOIN_TOKEN    → kubeadm join token for worker nodes
  JOIN_ENDPOINT → API server endpoint (IP:6443)

Switch context

ikctl --context staging

Run a kit

# Remote (default)
ikctl --install show-date --name workers

# With parameters passed to the script
ikctl --install deploy --name workers -p v2.1.0

# With sudo
ikctl --install install-docker --name workers --sudo sudo

# Local machine (no SSH)
ikctl --install show-date --mode local

# Preview without executing
ikctl --install deploy --name workers -p v2.1.0 --dry-run

Parallel execution

When a server group has multiple hosts, ikctl runs them concurrently:

ikctl --install show-date --name workers --parallel-workers 4

A spinner appears while connecting to each host and a progress bar during uploads. Output is prefixed with the host IP.

Pipeline orchestration

A pipeline chains kits across server groups with DAG dependencies and output passing:

# ~/kits/pipelines/install-kubernetes.yaml
name: install-kubernetes

steps:
  - id: prepare
    kit: packages
    servers: all-nodes
    sudo: true

  - id: init-master
    kit: kubeadm-init
    servers: master
    sudo: true
    needs: [prepare]
    params:
      - "{{ params.POD_CIDR }}"      # from CLI: -p POD_CIDR=192.168.0.0/16

  - id: join-workers
    kit: join-in-kubes
    servers: workers
    sudo: true
    needs: [init-master]
    params:
      - "{{ steps.init-master.JOIN_TOKEN }}"    # output from previous step
      - "{{ steps.init-master.JOIN_ENDPOINT }}" # output from previous step

# Run pipeline by name (requires path_pipelines in config)
ikctl --pipeline install-kubernetes -p POD_CIDR=192.168.0.0/16

# Or by full path
ikctl --pipeline ~/kits/pipelines/install-kubernetes.yaml -p POD_CIDR=192.168.0.0/16

# Preview
ikctl --pipeline install-kubernetes -p POD_CIDR=192.168.0.0/16 --dry-run

Steps without needs run in parallel. Steps fail gracefully — dependents are marked as skipped.

Debug mode

# Show all internal logs (paramiko included)
ikctl --install show-date --name workers --debug

Without --debug, only useful output is shown (Rich formatted).

Timeouts

ikctl --install show-date --name workers --timeout-connect 60
ikctl --install build --mode local --timeout-exec 300

Or set per context in ~/.ikctl/config with timeout_connect and timeout_exec.

License

Apache License 2.0 — see LICENSE for details.