Skaftor CLI MCP server
Commit Nov 12, 2026Drive AI-native software delivery work orders, manage execution context, and handle pull requests directly from your terminal via MCP.
Also known as skaftor/skaftor-cli
- Stars
- 1
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 12, 2026
- License
- MIT
What is Skaftor CLI MCP?
Skaftor CLI is an MCP server. Drive AI-native software delivery work orders, manage execution context, and handle pull requests directly from your terminal via MCP.
Summary generated from the project's README
About Skaftor CLI MCP server
Developers managing AI-driven software delivery can use this tool to coordinate work orders and manage execution context directly from their terminal. It acts as a command-line interface for the Skaftor system, serving as a thin wrapper over a project's MCP server. This ensures that the same work orders and governance seen by coding agents in tools like Cursor or Claude Code are available in the shell.
Users connect by providing a personal MCP link obtained from the Skaftor app via the login command. The tool requires Node.js 18 or higher and has zero external dependencies. It allows for listing repositories, tracking specific services, and interacting with work orders through commands like wo next, wo plan, and wo start. For premium organization plans, it also provides the ability to launch and manage cloud-based workstations using the up and ssh commands. It supports both cloud-hosted and self-hosted Operator setups via Kubernetes.
What you can do with it
- 1Manage software work ordersA developer can view their assigned tasks or the next recommended task to maintain momentum.
- 2Plan task executionYou can use the decision engine to analyze repositories, files, and potential impact before starting work.
- 3Launch cloud workstationsTeams with premium plans can spin up fully-loaded cloud environments with coding agents pre-installed.
Written from the project's README. Every use case is checked against a quote from it.
Automated checks
Ran Nov 15, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachableskaftor/skaftor-cli is public and reachablegithub.com/skaftor/skaftor-cli ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-12github.com/skaftor/skaftor-cli ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubSkaftor CLI
Drive AI-native software delivery from your terminal.
skaftor is a zero-dependency command-line client for Skaftor —
the system of record for AI-native software delivery. It's a thin wrapper over your
project's MCP server (the same surface coding agents like Cursor and Claude Code
connect to), so you can move work orders, pull execution context, and open pull
requests without leaving the shell.
Website · Docs · What is Skaftor? · Discussions
Why
When AI agents write most of the code, the bottleneck moves from typing to coordination, governance, and memory — the Orchestration Gap. Skaftor closes it: intent becomes agent-executable work orders, agents are orchestrated over MCP, every change is governed, and the whole history is remembered in an engineering memory graph. This CLI is the terminal front-end to that workflow.
Install
Repository mirrors use the same personal project connection as work orders:
skaftor repos ls --tier lazy --page 1
skaftor repos ls --q billing --provider gitlab # narrow a large organisation by name or code host
skaftor repos track github:my-team/my-service
Listings return up to 50 repositories per page; use the response's more field
and --page (1–100) to continue. --json preserves the MCP response, including
refusal reasons and nextAction. Refusals exit with status 1. The platform
checks your permissions, repository scope and organisation storage allowance.
Needs Node.js 18+. No other dependencies.
# via the hosted installer (drops `skaftor` into ~/.local/bin)
curl -fsSL https://app.skaftor.com/install | sh
# or via npm
npm install -g skaftor
Connect
Get your personal MCP link from the app: open a project → a work order → "connect your coding agent" and copy the URL. Then:
skaftor login "https://app.skaftor.com/api/mcp/<projectId>?dev=<token>"
skaftor context # list connections, ● marks the active one
skaftor whoami
The connection is saved to ~/.skaftor/platform.json (chmod 600 — it holds a token).
An older ~/.skaftor/config.json from this CLI is migrated once, automatically.
Work orders
skaftor wo next # the recommended next task for you
skaftor wo mine # everything assigned to you
skaftor wo list --status ready # filter by status or --phase
skaftor wo show AUTH-02 # full spec + acceptance criteria + manifest
skaftor wo plan AUTH-02 # decision engine: repo, files, reuse, impact
skaftor wo manifest AUTH-02 # the full execution manifest as one JSON payload
skaftor wo start AUTH-02 --branch feature/auth-2
skaftor wo ask AUTH-02 "Should partial reversals be idempotent?"
skaftor wo done AUTH-02 --pr https://github.com/acme/repo/pull/42 \
--summary "Implemented core path" --ide "Claude Code"
Full command reference: skaftor.com/docs/cli/reference.
Managed workstations (premium)
Launch a fully-loaded cloud workstation for a work order — coding agents pre-installed — and drive it from the same login. Available when the workstation feature is on your organisation's plan; otherwise every command below answers with a clear "not enabled — upgrade" message. The platform decides; this CLI holds no secret.
skaftor verify # preflight: plan entitlement, Skaftor Cloud connection, ready to launch
skaftor up --agent claude # launch a workstation on Skaftor Cloud (returns at once)
skaftor up --repo bitbucket.org/acme/app --branch feat/x
skaftor ws # your workstations and their status
skaftor ssh <name> # interactive shell (^D to exit)
skaftor stop <name> · skaftor start <name>
skaftor rm <name> --yes # delete it and its home volume
skaftor ssh uses a one-hour, single-workstation credential issued by the platform after
it checks your organisation's entitlement and that the workstation is yours; the CLI renews
it at ~45 min. If your organisation loses the feature, the renewal is refused and the
session ends at expiry. Your own login session simply re-authenticates after 24 h idle.
Operator (BYOC)
Run your organisation's data plane — content, credentials, runs — in your own Kubernetes cluster, with Skaftor Cloud
keeping sign-in, billing and orchestration. You need helm and kubectl (and gcloud for Secret Manager).
# 1. An Admin gets a one-time code in Settings → Operator
skaftor operator enroll <code> --address https://operator.example.com --kube-context my-cluster
# 2. Install (the version comes from Skaftor)
skaftor operator install --tag <version> --chart-version <chart version> --kube-context my-cluster \
--host operator.example.com --tls-secret operator-tls --gsa skaftor-operator@my-project.iam.gserviceaccount.com
# Later
skaftor operator upgrade --tag <version> --kube-context my-cluster
skaftor operator status --kube-context my-cluster
Always name the cluster context (--kube-context, with --kubeconfig if it lives in another file): the CLI never
uses whichever context happens to be current. One machine keeps one Operator's state (~/.skaftor/operator.json,
nothing secret); for a second, set SKAFTOR_OPERATOR_STATE to its own file. The Operator's key and credentials key stay with you — in a Kubernetes Secret, or with `--store gcp --project
How it works
The CLI speaks JSON-RPC 2.0 to your project's MCP server at
<origin>/api/mcp/<projectId>?dev=<token> — the identical surface coding agents use.
That means the CLI, Cursor, and Claude Code all see the same work orders, the same
execution context, and the same governance. Point it at a self-hosted instance with
SKAFTOR_URL=https://your-host.
Contributing
Issues and PRs welcome — see CONTRIBUTING.md and our
Code of Conduct. Good first issues are labeled
good first issue.
Security
Found a vulnerability? Please follow our Security Policy — do not open a public issue for security reports.
License
MIT © Skaftor
Skaftor CLI MCP questions
How do I connect the CLI to my project?
You can connect by using the login command with your personal MCP link obtained from the Skaftor app.
What are the system requirements for Skaftor CLI?
The CLI requires Node.js version 18 or higher and has zero other dependencies.
Can I run Skaftor in my own infrastructure?
Yes, you can run the Skaftor Operator in your own Kubernetes cluster to manage your organization's data plane.
Does the CLI support cloud workstations?
Yes, premium users can launch fully-loaded cloud workstations for work orders using the 'skaftor up' command.