About shipeasy
shipeasy is an experimentation and release-management platform for coding agents, using natural-language workflows for flags, configs, experiments, metrics, and translations. It combines an MCP server with agent skills and plugins to support setup from configuration through code changes.
Key features
- 01Manage feature flags and kill switches
- 02Configure dynamic settings and A/B tests
- 03Track event metrics
- 04Manage translations
- 05Connect SDK setup with code changes through agent workflows
- 06Collect feedback, errors, and alerts
Use cases
- 01Roll out a new feature gradually
- 02Add a kill switch for checkout
- 03Compare signup button experiment results
- 04Triage user feedback and production alerts
MCP support
The shipeasy MCP server exposes platform tools to coding agents; skills and plugins work with it to support workflows such as configuring flags, settings, and experiments. Agents connect through plugins or shared skills and MCP configuration.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://mcp.shipeasy.ai/mcpclaude mcp add --transport http ai-shipeasy https://mcp.shipeasy.ai/mcp{"mcpServers":{"ai-shipeasy":{"type":"http","url":"https://mcp.shipeasy.ai/mcp"}}}Open in Cursor{"servers":{"ai-shipeasy":{"type":"http","url":"https://mcp.shipeasy.ai/mcp"}}}Open in VS Code{"mcpServers":{"ai-shipeasy":{"type":"http","url":"https://mcp.shipeasy.ai/mcp"}}}{"mcpServers":{"ai-shipeasy":{"serverUrl":"https://mcp.shipeasy.ai/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
115| Parameter | Type | Required | Description |
|---|---|---|---|
| bind | boolean | No | Write .shipeasy after upsert. Default true. Set false to skip binding. |
| name | string | No | Human-readable project name. Defaults to the domain on first create; ignored on later upserts. |
| path | string | No | Directory to write .shipeasy in. Defaults to the MCP server's cwd. |
| domain | string | Yes | Hostname-like identifier for the project (e.g. shouks.app, acme.com). Primary key for upsert. |
Details
From the project README
Feature flags · kill switches · dynamic configs · A/B experiments · metrics · i18n — spun up from a single sentence to your AI coding agent.
What is Shipeasy?
Shipeasy is the AI-native experimentation platform. Feature flags, kill switches, dynamic configs, A/B tests, metrics and translations — all controllable from where you already work: a sentence to your AI coding agent.
"Ship the new pricing page behind a flag at 5%." "Add a kill switch for checkout in case it breaks." "Run an A/B test on the signup button and tell me which wins."
No dashboards to learn, no SDK boilerplate to hand-write. Shipeasy installs as a plugin into your coding agent and an MCP server, so Claude Code (and 70+ other agents) can create the flag, instrument the metric, wire the SDK call, and roll it out — end to end, from plain English.
This directory is the source-of-truth for that plugin marketplace.
One plugin tree, many hosts — the same shipeasy/skills/ directory and the
same shipeasy MCP server feed every coding agent below. Each host gets a tiny
manifest that points at the shared files; nothing is duplicated.
Why teams use it
- 🚀 From one sentence to shipped. Describe what you want; the agent creates the gate/config/experiment, instruments the event, and edits the call site.
- 🎛️ The whole release toolkit. Feature flags, kill switches, dynamic configs, A/B experiments, event metrics and i18n — one SDK, one key per entrypoint, one install.
- 📊 Statistics that won't lie to you. Sequential testing, SRM detection, auto-ramp guardrails and holdouts come built in — no stats degree required.
- ⚡ Edge-fast, globally. Built on Cloudflare Workers + KV with infinite-TTL CDN caching and explicit purge — flag reads never touch a database on the hot path.
- 🤖 Works with your agent, not against it. Native plugins for Claude Code, Codex and Copilot CLI; shared skills + MCP for everything else.
- 🛟 Closes the loop. An operational inbox turns bug reports, feature requests, production errors and metric alerts into one queue your agent can burn down — even on a schedule, unattended.
Install in 12 seconds
npm install -g @shipeasy/cli
shipeasy setup
That's it. shipeasy setup detects your coding agent, installs the right
integration for it (native plugin where one exists, shared skills + MCP
everywhere else), authenticates, binds the repo to a project, mints keys and
installs @shipeasy/sdk. One command, every host — then just talk to your agent.
Prefer to wire the agent integration yourself? Per-host install commands
There are two install tiers. Tier 1 hosts have a native plugin system, so
one command bundles skills + MCP. Tier 2 hosts (70+ agents) take the same
shared skills via the
vercel-labs/skills CLI plus a small
MCP config snippet. (shipeasy setup above does the right one for you.)
Tier 1 — one-command plugin install
| Agent | Install |
|---|---|
| Claude Code | claude plugin marketplace add shipeasy-ai/shipeasy → claude plugin install shipeasy@shipeasy |
| Codex | codex plugin marketplace add shipeasy-ai/shipeasy → codex plugin add shipeasy@shipeasy (or, in the TUI, /plugin marketplace add shipeasy-ai/shipeasy → /plugin add shipeasy@shipeasy) |
| GitHub Copilot CLI | copilot plugin marketplace add shipeasy-ai/shipeasy → copilot plugin install shipeasy@shipeasy |
Tier 2 — skills + MCP (Cursor, Windsurf, Cline, Gemini, OpenCode, Continue, …)
# 1. skills — installs all 17 shipeasy-* skills
npx skills add https://github.com/shipeasy-ai/shipeasy -a <agent>
# 2. MCP — add the shipeasy server to that agent's MCP config:
# { "mcpServers": { "shipeasy": { "command": "npx", "args": ["-y", "@shipeasy/mcp@latest"] } } }
<agent> is e.g. opencode, cursor, windsurf, cline, gemini-cli,
continue, github-copilot. The full per-host list, exact MCP file paths and
formats live in INSTALL.md (also published at
https://docs.shipeasy.ai/get-started/agents).
Then: "set up shipeasy"
Installing the plugin registers the skills and MCP tools — it does not run any shell commands. To go live:
- Run
shipeasy setup(or say "set up shipeasy") — one command detects your subprojects, installs@shipeasy/sdk, authenticates, binds the repo to a project, mints keys, installs the SDK + feature how-to skills into your agents, offers the modules + devtools overlay, and emitsshipeasy-wiring.mdwith the remaining code steps. - Turn on the modules you want —
shipeasy setupoffers them, or enable one directly withshipeasy install <group>:shipeasy install flags— flags + configs + kill switches + experiments + events (the whole decide-at-runtime platform, one pass).shipeasy install ops— feedback (bugs + feature requests) + production errors + alerts.shipeasy install i18n— translations.
The matching per-area how-to skills (shipeasy-flags, shipeasy-ops,
shipeasy-i18n) are installed alongside. After that, just talk to your agent.
You ▸ Ship a feature gate for the new pricing page, rolled out to 5% of users.
Agent ▸ Created gate `new_pricing_page` (rollout 5%). Add gates.check("new_pricing_page")
at apps/web/app/pricing/page.tsx:12 — done. Want me to bump it to 25%?
Works with your agent
| Native plugin (Tier 1) | Claude Code · Codex · GitHub Copilot CLI |
| Skills + MCP (Tier 2) | Cursor · Windsurf · Cline · Gemini CLI · OpenCode · Continue · OpenClaw · 70+ more |
All 11 skills (9 area guides + 2 multi-step workflow skills) and the
shipeasy MCP server port to every agent — the flows trigger from
natural-language phrasing everywhere, and Claude Code additionally exposes each
skill as a /shipeasy:<skill-name> command. Skills and MCP are referenced,
never copied.
Learn more
- 📖 Skill reference — every skill, its auto-triggers, MCP tool and CLI mapping, plus the headline workflows.
- 🧩
INSTALL.md— full per-agent install reference (70+ hosts). - ⏰ Trigger install guide —
shipeasy ops trigger create <provider> --help(also at https://docs.shipeasy.ai/get-started/triggers) — set up unattended scheduled triggers that burn down your ops inbox and open one PR per item. - 🌐 shipeasy.ai · docs.shipeasy.ai
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.