For builders shipping with Claude Code, Gemini, and Cursor

Faster feedback loop
for AI coding agents.

Capture a browser bug with one click. Hand it to Claude Code, Gemini, Cursor, or any MCP-compatible agent - with the screenshot, console errors, network logs, DOM context, and user events the agent needs to ship a real fix.

Start Free See How It Works

Free for solo developers

Chrome extension capture

CLI + MCP integrations

Claude

Codex

Cursor

Gemini

OpenClaw

Hermes

Antigravity

Kiro

Claude

Codex

Cursor

Gemini

OpenClaw

Hermes

Antigravity

Kiro

Capture

Analyze

Fix

01 · Browser Capture

Click elements to pin

ESC

1 pin

ESC

Checkout

Confirm your order

Pro Plan · Annual 12 months billing cycle

$240.00

Card number

•••• •••• •••• ____

Expiry

12/27

CVC

•••

TypeError

payment.handler is undefined

disbug-7392.json

Download complete

02 · AI Agent

Claude Code › disbug-mcp

v4.5 Opus

Connected

One click in the browser, full evidence package in your agent.

Same session. Same data. Across the CLI and MCP tools.

The Problem

Your AI agent can't fix
what it can't see.

You paste a screenshot. You paste the console output. You type reproduction steps from memory. The agent makes a half-context guess. Two iterations later, it still hasn't seen the failing network call. You give up and fix it yourself.

Manual Reproduction

The Disbug Way

One click.
Full context. Real fixes.

Click the Disbug toolbar icon. Disbug captures the screenshot, console errors, network logs, DOM selector, and user-event trail. Claude Code, Codex, or Cursor reads the full evidence package over MCP and ships the fix on the first try.

Instant Ground Truth

From browser bug to AI fix,
in four steps.

Capture with one click

Click the pinned Disbug icon on any web app. The extension grabs the screenshot, DOM selector, console errors, failed network requests, user-event trail, browser metadata, and local storage all in the background.

Save the session to your project

Save the captured session to the right Disbug project so your dashboard, CLI, and coding agent all work from the same session.

Wire it to your agent

Run disbug mcp, or add Disbug to your Claude Desktop / Cursor / Codex / OpenClaw / Hermes config once. Your agent gets MCP tools to pull the latest session, search past pins, and fetch only the fields it needs.

Agent reads, agent fixes

Claude Code or Codex pulls the screenshot, the console error, and the failing request. It identifies the bug, writes the fix, runs it, and opens the PR. You go from browser bug to merged commit in minutes not hours.

Cloud Workflow / v1.0

Instant agent pickup.
Zero manual context copying.

Sync browser evidence for immediate discovery. Give your AI agent the structured context it needs to ship the fix on the first try.

01 · Browser Capture (Cloud)

Click elements to pin

ESC

1 pin

ESC

Saved to Cloud

Checkout

Confirm your order

Pro Plan · Annual 12 months billing cycle

$240.00

Card number

•••• •••• •••• ____

Expiry

12/27

CVC

•••

TypeError

payment.handler is undefined

Save session

Session Title

Project

QA Team

Frontend Project

Pins 1 total

Pay button error

02 · AI Agent

Claude Code › disbug-cloud

Cloud MCP

Live Sync

disbug watch --cloud-only

watching: cloud (since=none, poll=30s)...

Built for the way agents actually consume context

Automating the flow of context from browser to agent.

app.acme.com/checkout

Checkout

Confirm your order

Pro Plan · Annual 12 months

$240.00

Card number

•••• •••• •••• ____

Expiry

12/27

CVC

•••

Uncaught TypeError: payment.handler is undefined

Disbug

Capturing

Session #7392

Console 3 errors

Network 2 failed

Events 14

Replay 00:42

claude_desktop_config.json

MCP

~/.config/claude/claude_desktop_config.json

Add once

{
  "mcpServers": {
    "disbug": {
      "command": "disbug",
      "args": ["mcp"]
    }
  }
}

Connected

8 tools available

Same config works for

Claude Code Claude Desktop Cursor OpenClaw

claude code — disbug-mcp

MCP

Claude Code · disbug-mcp

MCP

Transparency / v1.0

Built for builders.
Honest about it.

Disbug is in active development by a founder who ships with Claude Code and Codex every day. Here's what's true about the product today.

Module 01

Open Protocol

MCP-native, not proprietary

Built on the open Model Context Protocol. Works with any compliant agent today and any compliant agent that ships tomorrow. No lock-in to a single AI vendor.

Module 02

Scoped Agent Access

Your sessions stay controlled

Agent access is read-only and scoped to the teams and projects you authorize. Revoke an agent whenever its access is no longer needed.

Module 03

Build in Public

Built in the open

Disbug is being rebuilt for the AI coding era in public. Roadmap, releases, and lessons from the rebuild are posted regularly follow along on X at @aswin_db and on GitHub.

Principles / v1.0

Why builders ship with Disbug
The evidence standard.

Agent-readable by design

Console errors, network failures, DOM selectors, and user-event trails structured exactly how Claude Code and Codex expect to consume them.

One session, shared context

Capture once, then open the same session from the dashboard, CLI, or MCP tools without rebuilding the debugging context.

Selective fetching

Your agent pulls only what it needs console logs OR screenshot OR network instead of dumping the whole bundle into context.

Built for the AI coding workflow

One-command setup for Claude Code, Codex, Cursor, and more. No SDK to learn. Just a binary and an MCP entry in your config.

Use Cases / v1.0

Built for the people shipping with agents
Scale your visibility.

Solo AI builders

You ship with Claude Code or Codex daily. Hit a browser bug, capture everything in one click, and hand the full session to your agent.

Indie · Vibe coding · Free

AI-first product teams

Your team's coding agents are your most productive 'engineers.' Give them the browser evidence they're missing.

2–5 people · Pro tier

QA-led product teams

QA captures bugs. Developers' AI agents pull the full evidence package automatically. The handoff takes 30 seconds.

QA · Enterprise · Paid

Digital agencies

Client-side feedback, dev-side AI fixes. Per-project workspaces and scoped agent tokens keep client work separated.

Multi-client · Paid

Ready to ship faster?

Start in 30 seconds.

Install the extension. Add the MCP server to your agent config. Watch Claude Code or Codex actually fix the bug instead of guessing at it.

Start Free See How It Works

Free for solo developers

CLI + MCP integrations

Faster feedback loop for AI coding agents.

One click in the browser, full evidence package in your agent.

Your AI agent can't fix what it can't see.

One click. Full context. Real fixes.

From browser bug to AI fix, in four steps.

Capture with one click

Save the session to your project

Wire it to your agent

Agent reads, agent fixes

Instant agent pickup. Zero manual context copying.

Built for the way agents actually consume context

Built for builders. Honest about it.

MCP-native, not proprietary

Your sessions stay controlled

Built in the open

Why builders ship with Disbug The evidence standard.

Agent-readable by design

One session, shared context

Selective fetching

Built for the AI coding workflow

Built for the people shipping with agents Scale your visibility.

Solo AI builders

AI-first product teams

QA-led product teams

Digital agencies

Start in 30 seconds.

Faster feedback loop
for AI coding agents.

Your AI agent can't fix
what it can't see.

One click.
Full context. Real fixes.

From browser bug to AI fix,
in four steps.

Instant agent pickup.
Zero manual context copying.

Built for builders.
Honest about it.

Why builders ship with Disbug
The evidence standard.

Built for the people shipping with agents
Scale your visibility.