AI agent activity log

The activity log for AI work inside your app.

By end of 2026, AI agents will do 30–60% of meaningful work inside production apps. Nobody has a clean way to see that work happening, explain it to non-technical colleagues, or let the AI itself answer "what did I do yesterday?". EZLogs is built for exactly that.

What it does

The EZLogs agent drops into your Rails or Next.js app and captures every HTTP request, every background job, and every database change — whether the actor is a human user, a system process, or an AI agent. EZLogs correlates those events into a single “action” per business event, then renders each action as a plain-English card:

Every claim links back to the underlying action. Every citation is real. The AI never invents data — it only phrases events your app actually emitted.

Two readers, one source of truth

For your team — Pulse pages, in plain English

Open /actors/agent:claude or /entities/Order/4471. Live state. Sentence-level summaries. Citations on every value. The same screen support uses to answer “what happened to this order?” and product uses to ask “what is this agent doing?”

For your AI — MCP server + in-app chat

Six query-only MCP tools at POST /mcp for Claude Desktop / Cursor / ChatGPT. Plus an in-app “Ask EZLogs” chat sidebar that runs the same tools in-process. Your AI finally has memory of what your app did, and answers your team in their own words with citations they can click through.

Read-only by design

EZLogs reads activity from your app. It never writes back. There are no “pause agent” toggles that throttle your code, no approval queues that send anything anywhere, no MCP tools that perform actions. The MCP layer is for asking what happened, not making something happen. If you compromise an EZLogs account, the blast radius is read-only.

How is this different from agent observability tools?

Tools like LangSmith, Langfuse, Arize, Braintrust are built for the engineer debugging the agent — token counts, latencies, eval scores, trace inspection. EZLogs is built for the rest of the company watching the agent do work in production. We don’t compete with those tools. We’re the layer above them — the activity log your PM, founder, and support team can read.

Drop-in install

One gem (Rails) or one npm package (Next.js). Zero per-route code, zero per-agent instrumentation.

Rails

bundle add ez_logs_agent

# config/initializers/ezlogs.rb
EzLogs.configure do |c|
  c.api_key = ENV["EZLOGS_API_KEY"]
end

Next.js

npm install ezlogs-nextjs

// next.config.ts
import { withEzlogsConfig } from "ezlogs-nextjs/plugin";
export default withEzlogsConfig({});

Restart. Every HTTP request, every job, every DB change — including everything your AI agents do — starts flowing into EZLogs immediately.