EZLogs over MCP

Ask Claude, ChatGPT or Cursor what happened in your app

EZLogs turns the requests, background jobs and database changes in your Rails or Next.js app into plain-English actions. The EZLogs MCP server lets the AI assistant you already use read that log, so you can ask it what happened, who did it and what failed. It can only read. It cannot change anything in your app.

Server URL: https://app.ezlogs.io/mcp
Sign in with your EZLogs account (OAuth). Requires the EZLogs agent installed in a Rails or Next.js app. Free during the beta.

What your AI can see

Example: a night of ferry rebookings

These answers come from a demo account: Brackwater Ferries, a fictional ferry company where an AI rebooking assistant handles disrupted bookings. A gale cancelled two sailings overnight.

Claude answering 'what did the rebooking assistant do last night?' using EZLogs: 48 bookings disrupted, 40 rebooked, 5 refunds issued, two sailing-full rejections that went through on a retry, and a table of the three failed refunds totalling 152.00 EUR
Claude, connected to EZLogs over MCP, answering a question about the demo account.

“What did the rebooking assistant do last night? Anything I need to deal with?”

Two sailings were cancelled at about 04:26, citing the gale warning, which disrupted 48 bookings. The Rebooking assistant moved 40 onto other sailings and refunded the other 8. Five refunds went through (€355.45). Three failed because the card had expired, and those bookings are now in “refund failed”.

“What happened to refund RF-M9PM?”

It failed at 03:27 UTC. The Rebooking assistant tried to refund €64.20 for booking BRK-M9PM. The refund went from pending to failed because the card had expired, and the booking was marked “refund failed”.

“Which refunds failed, and how much money is stuck?”

Three: RF-M9PM (€64.20), RF-9LDK (€55.70) and RF-3JRW (€32.10), €152.00 in total. Nothing in the log shows anyone contacting these passengers yet.

“Were any rebookings rejected?”

Two were rejected as “sailing full” (BRK-QSKS and BRK-Y4CZ). Both went through on a retry.

Other questions that work well: “What did sarah@example.com do today?”, “Show me failed checkouts this week”, “Who were the most active users in the last 7 days?”, “Why did this order succeed and that one fail?”.

Connect your AI assistant

Every client uses the same URL, https://app.ezlogs.io/mcp, and signs in through your browser. No API key to copy.

Claude (claude.ai and Claude Desktop)

  1. Open Customize → Connectors → Add custom connector.
  2. Name it EZLogs and paste https://app.ezlogs.io/mcp.
  3. Select Connect and sign in to EZLogs.

On Team and Enterprise plans an owner may need to add the connector for the organisation first.

Claude Code

claude mcp add --transport http ezlogs https://app.ezlogs.io/mcp

Then run /mcp inside Claude Code and choose ezlogs to sign in.

ChatGPT

Custom MCP connectors need a paid ChatGPT plan, and on Business or Enterprise an admin may need to allow them. Turn on Developer mode in Settings (under Apps → Advanced), then create a connector with the URL https://app.ezlogs.io/mcp and OAuth authentication. ChatGPT moves these menus from time to time; the connector itself stays the same.

Cursor

Open Cursor’s Customize page and add a new MCP server with the URL. Or add this to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "ezlogs": { "url": "https://app.ezlogs.io/mcp" }
  }
}

Cursor asks you to sign in the first time.

Codex

codex mcp add ezlogs --url https://app.ezlogs.io/mcp
codex mcp login ezlogs

Lovable

Open Connectors → Add custom → MCP server, paste the URL, keep OAuth, then add and authorize. If MCP server is greyed out, a workspace admin has turned custom servers off.

Tools (all read-only)

Every tool is marked readOnlyHint: true. None of them can create, change or delete anything, in EZLogs or in your app.

ToolWhat it does
find_actionsSearch actions by text, actor, entity, outcome (success, failure, partial failure), intent and date. Answers “what happened” questions.
get_actionOpen one action with its full list of events (requests, jobs, database changes).
actor_timelineOne person’s or agent’s recent actions, plus their current state with links to the actions behind each value.
entity_timelineOne record’s recent actions (an order, a booking, a project), plus its current state.
top_listsThe most active actors and entities over a window of up to 90 days.
compare_actionsCompare 2 to 5 actions and show where they first differ. Useful when one worked and one failed.

Security and privacy

FAQ

Can my AI assistant change anything in my app through EZLogs?

No. All six tools only read your activity log. EZLogs has no write access to your app, and the MCP server has no tools that create, update or delete anything.

Do I need an EZLogs account?

Yes. The MCP server reads the activity log EZLogs collects from your app, so you need an EZLogs account (free during the beta) with the agent installed in a Rails or Next.js app. Without it there is nothing to query.

Which AI clients work?

Any client that supports remote MCP servers over HTTP with OAuth and dynamic client registration. We have setup steps above for Claude (claude.ai and Claude Desktop), Claude Code, ChatGPT, Cursor, Codex and Lovable.

Who can see my data?

Only people in your EZLogs company, the AI clients they sign in, and the subprocessors on our Subprocessors page (hosting in Germany, and the AI providers used for optional explanations). Each connection is tied to one user, can be revoked from Settings → Connectors → MCP, and uses short-lived access tokens. Data is stored in Germany (EU).

Does EZLogs use AI to make up the answers?

No. The tools return actions built from the real events your app sent: requests, background jobs and database changes. Your AI assistant writes the answer from that data, and each action has an id you can open in EZLogs to check it.

Start free →

No card required · Read-only by design

← Back to home