Everything the CLI does, then it keeps watching.
Skene Cloud builds the same journey the CLI builds, then adds what a one-shot run cannot: review on every release, the journey canvas, launch checks, Flows, and a hosted MCP server your coding agent can call.
Sign in, create a workspace, connect a source. The free tier is one audit, no card.
Starts with a free audit
Sign in, connect a source, one audit per workspace. No card, no trial clock.
Never pushes to your default branch
Everything Skene proposes arrives as a pull request you review and merge yourself.
Your product data stays in your own database
Supabase connects over OAuth and starts read-only. Row data is never read.
The CLI's journey, and everything around it.
The journey.yaml the CLI writes is the same artifact Cloud builds and keeps. The difference is what happens after the run.
- Plan
- Verify
- Launch ready
- Check
A journey.yaml you push from the CLI outranks the Analyze pipeline: a hand-maintained journey always wins.
| Capability | Skene CLI (OSS) | Skene Cloud |
|---|---|---|
| Journey from code and schema, seven stages | ✓ | ✓ |
| PostgreSQL connector, credentials never stored | ✓ --db-url | ✓ Supabase OAuth, read-only |
| Runs on your machine, your model provider, MIT | ✓ | Managed |
| Review on every release, one-click fixes, /skene fix | – | ✓ |
| Journey canvas, editable, agent-assisted | – | ✓ |
| Launch checks: plan, verify, launch ready, check | – | ✓ |
| Events table with statuses and ignore rules | – | ✓ |
| Flows: one script tag, real user paths, 90-day retention | – | ✓ |
| Hosted MCP server, 11 tools for coding agents | – | ✓ |
| Record kept between runs | – | ✓ |
The release check, in your review flow.
Skene reviews every pull request on the repository you link for removed or broken tracking. It requests changes with inline comments on the exact lines, or passes.
- One-line fixes arrive as committable GitHub suggestions
- Anything larger arrives as a “Prompt to fix this” block, pasteable into a coding agent
/skene fixopens a follow-up PR restoring what the diff deleted, and only that- When a later run passes, Skene dismisses its own earlier change requests
/skene fixThe workspace: canvas, events, Flows.
The journey lives on an editable canvas. Describe a change in plain language and the agent proposes it as a ghost node; accepting writes it into journey.yaml. Skene turns a hypothesis into a plan, and you never need to supply event names.
- Verify compares required events against your code and arriving evidence: ready or missing
- Check computes the metric after launch: current, previous, change, goal met or not
- Flows maps real user paths from one dependency-free script tag, purged after 90 days
- Plan
- Verify
- Launch ready
- Check
For your agent: 11 tools over MCP.
The hosted MCP server authenticates with a workspace API key. Read tools need read scopes; in read-only mode write tools are filtered out of the list entirely. The deterministic gate is callable over HTTP or as an MCP tool, and it can disagree with the model review by design.
- Two key types: a server-side secret key, and a write-only publishable key safe to embed
- Exactly two roles, owner and member. No permission matrix to configure
- Lifecycle
- analyse, build, get, view journey
- Events
- tracking events, check, gap, ack, analytics review
- Launch check
- feature check
Start with the free audit.
One audit per workspace, once. No card.