MCP
SnowAnalytics has an MCP server. Connect it to Claude, Cursor, VS Code, ChatGPT or any other MCP client and an agent can read your stats, errors, feedback and recordings, install SnowAnalytics on a site, and fix a bug in your code and check in production that the fix worked.
The endpoint is https://mcp.snowanalytics.app/mcp. It works on every plan.
What an agent can do
- Read. Overview numbers, any metric by any dimension, what changed between two periods, ranked error issues with the stack and what visitors did, visitor feedback with screenshots, and session recordings as text.
- Set up. Add a website, get the install steps for its framework with the features that fit turned on, and check that visits are arriving.
- Fix. Read an error, change your code, mark the issue fixed and let SnowAnalytics confirm in production. See Fix errors with an agent.
- Do the rest of the dashboard. Goals, funnels, alert rules, tags, chart notes, settings and more, through a handful of tools.
Access follows your account
Signing in gives the agent your account, not one site. It reaches every site you can open, across all your workspaces, with the role you have in each one. If you can't see a site in the dashboard, neither can the agent. Someone with only some sites in a workspace gives the agent only those.
When you connect, the consent screen asks two things. Connected apps then lists the connection as Read & act or Read only.
| Choice | What it means |
|---|---|
| Make changes for you | On, the agent can read and also change what you could change in the dashboard, such as resolving issues or adding goals. Off, it can only read, and anything that changes data is refused. |
| Visitor names & emails | Off by default. When off, identified visitors come back without names or emails. |
You can disconnect from Account → Connected apps at any time. See Security.
What stays in the dashboard
Some things only a person should do, even when asked. The tool answers with the dashboard link instead.
- Deleting, resetting or transferring a site, and making a dashboard public.
- Billing changes. An agent can fetch a checkout or billing portal link for you to open.
- Team members and their roles.
- API key secrets and your sign-in security.
- Sending email to visitors. An agent writes a draft and you send it.
- Features that run SnowAnalytics' own AI on its key, such as the error cause and asking a recording a question. The agent reads the same data directly instead.
The reason is simple: each of these either costs money, exposes data to someone new, or speaks to your visitors in your name. See Security for the full reasoning.
Prefer a terminal? The CLI runs the same tools as commands, for you, your scripts and CI.