MCP server
JourneyKit runs a remote Model Context Protocol server. Add it to your AI assistant and you can work with your workspace in plain words, "how did activation do last week?", "build a segment of trial contacts who never invited a teammate", "build a workflow that nudges them after three days".
https://journeykit.io/api/mcpCoding agents: install the plugin
If you want your coding agent to set JourneyKit up in your product, install the JourneyKit plugin instead of adding the server by hand. It bundles this MCP server with skills that teach the agent the whole setup. Its onboard skill reads your codebase and:
- defines your contact fields and event types in JourneyKit, from your contact model and what contacts do in your product;
- adds the SDK (
@journeykit.io/browser-sdkin JavaScript projects, the ingest API from other backends) and the identify and track calls; - matches your email branding to your product, writes email presets, and builds workflows that react to your events;
- checks that data arrives.
Workflows your assistant creates start paused, so nothing reaches your contacts until you turn them on. In Claude Code, one command installs the plugin and starts onboarding:
claude plugin marketplace add JourneyKit/journeykit-plugin && claude plugin install journeykit@journeykit && claude "/journeykit:onboard"In other agents, install the plugin and say "Onboard this project to JourneyKit." The dashboard's Get started page has the commands for Claude Code, Codex, Cursor and other agents, and shows the agent's progress live.
Signing in
The server uses OAuth. The first time your assistant connects, it opens JourneyKit in your browser: sign in, pick the workspace it may use, and allow access. You never copy an API key into your assistant.
The assistant acts as you, in that one workspace: it can see and change what you can there. Changes it makes are recorded as yours. You can see and disconnect every connected app under Settings → Connected apps; disconnecting takes effect immediately.
For scripts and agents that can't open a browser, send a secret API key instead: Authorization: Bearer jk_live_…, created under Settings → API keys.
Tools
The server has one tool for each operation of the REST API, named after it: contacts_list, segments_preview, workflows_create, broadcasts_send. Every tool is the same operation the dashboard and the REST API use, with the same validation.
Tools that only read are marked read-only, so assistants can usually run them without asking. Tools that change anything are marked destructive, so assistants ask first. Workflows your assistant creates through the plugin start paused.
Two tools exist for installing JourneyKit in a codebase:
| Tool | What it does |
|---|---|
check_setup | The SDK and ingest URLs, live API keys (prefixes only), what has arrived so far, and a checklist of what's missing. With since, it reports only what arrived after that time, how an agent verifies its install. With contactId, it looks up one contact. |
create_public_key | Creates a public jk_pub_ key for the browser SDK. Secret jk_live_ keys are never created over MCP: you create one under Settings → API keys and put it in your backend's environment yourself. |
Not available over MCP: API keys, billing, and support messages.
Set up your client
Claude Code
claude mcp add --transport http journeykit https://journeykit.io/api/mcpThen run /mcp in Claude Code and sign in.
Claude
In Claude (web or desktop), open Settings → Connectors → Add custom connector. Name it JourneyKit, enter https://journeykit.io/api/mcp, and connect.
ChatGPT
Open Settings → Connectors, turn on developer mode, and create a connector named JourneyKit with the server URL https://journeykit.io/api/mcp.
Cursor
Add to ~/.cursor/mcp.json (or .cursor/mcp.json in a project), then sign in from Cursor's MCP settings:
{
"mcpServers": {
"journeykit": { "url": "https://journeykit.io/api/mcp" }
}
}VS Code
Add to .vscode/mcp.json, then start the server from the file and sign in:
{
"servers": {
"journeykit": { "type": "http", "url": "https://journeykit.io/api/mcp" }
}
}Codex
codex mcp add journeykit --url https://journeykit.io/api/mcpOther clients
Clients that only run local servers can use the mcp-remote bridge:
{
"mcpServers": {
"journeykit": { "command": "npx", "args": ["mcp-remote", "https://journeykit.io/api/mcp"] }
}
}Access
Clients request two scopes: journeykit:read to look things up and journeykit:write to make changes. A client connected with read access only is asked to re-authorize the first time it tries to change something.