Install BugMojo in Cursor
One click from the install deeplink, or a four-line mcp.json, plus a rule that only loads when you are in test code.
Before you start
- Cursor installed.
- A BugMojo account you can sign in to in a browser.
- At least one captured bug to test against. The connection inherits your role, so Cursor sees only the bugs you can already see.
One click
Opens Cursor and installs the hosted BugMojo server. If nothing happens, copy the link below into your browser.
cursor://anysphere.cursor-deeplink/mcp/install?name=bugmojo&config=eyJ1cmwiOiJodHRwczovL3d3dy5idWdtb2pvLmNvbS9hcGkvbWNwP21vZGU9cWEifQ%3D%3DThe link installs the hosted endpoint with the QA tool surface. Cursor then prompts for the browser login on the first tool call.
Or by hand
Create
.cursor/mcp.jsonat your repo root for one project, or~/.cursor/mcp.jsonfor every project, and paste the config below.Fully quit and reopen Cursor.
Open Customize in Cursor's sidebar and find bugmojo. When it asks you to sign in, complete the BugMojo login in the browser and choose the one workspace the agent may use.
Back in Cursor, check that bugmojo is enabled and connected.
{
"mcpServers": {
"bugmojo": {
"url": "https://www.bugmojo.com/api/mcp?mode=qa"
}
}
}Try it
In Cursor's chat, name a bug by its key or paste its BugMojo share link, and ask something like "What broke in this bug?" Cursor calls the BugMojo tools, reads the replay summary, console and network logs, and suggests a fix grounded in what happened. See the MCP tool reference for every tool it can call.
What the plugin adds
| Part | What it does |
|---|---|
mcp.json | One server: the hosted endpoint, OAuth in the browser. |
rules/bugmojo.mdc | A rule globbed to tests/**, e2e/** and playwright.config.*. It enters context only when you are in test code. |
skills/ | The same skills the Claude Code plugin ships: plan, generate, run, assign, report, fix a bug, triage. |
The rule teaches Cursor how to annotate a spec so a result maps to a test case, and which flakiness patterns to avoid: unordered-collection assertions, hard waits, missing assertions, network-timing assumptions.
Troubleshooting
- The server does not connect or does not appearLikely cause: Cursor was not fully restarted, the JSON does not parse, or the URL has a typo.Quit Cursor entirely (not just the window), reopen, and confirm the URL is exactly
https://www.bugmojo.com/api/mcpwith any query string after it. Cursor's MCP log shows the reason: open the Output panel and choose MCP Logs. - The deeplink opens Cursor and installs nothingLikely cause: The base64 payload encodes the wrong object.The payload must be the server object (
{"url":"…"}), not the wholemcpServersmap. Generate the link rather than hand-encoding it. - Tools are missing that you expectedLikely cause: Your role lacks the permission, or the mode switch is not set.Call
whoamito see your role and permissions, or reconnect with?mode=all. - The rule never firesLikely cause: You are not editing a file it globs.It is deliberately scoped to test files. Open something under
tests/ore2e/, or ask about the conventions directly. - The connection is refusedLikely cause: AI agents and MCP are turned off for the workspace.An owner or admin can turn AI agents & MCP back on under Settings, Security. While it is off, every MCP client is refused. Still stuck? Contact support.
Common questions
Which Cursor version do I need?
3.13.0 or later. The plugin manifest declares that as its minimum client version.
Why is the rule not always on?
It is globbed to <code>tests/**</code>, <code>e2e/**</code> and <code>playwright.config.*</code>. A rule that loads on every request is a tax on every request.
The deeplink opened Cursor but installed nothing.
The base64 payload must be the server object alone. Generate the link with the plugin's deeplink script rather than hand-encoding it.
How is this different from just adding the MCP server?
The MCP server gives Cursor the tools. The plugin adds the rule and the shared skills: planning, generating, running, assigning and reporting.
Where does Cursor store the MCP config?
In an <code>mcp.json</code> file. Use <code>~/.cursor/mcp.json</code> for every project, or <code>.cursor/mcp.json</code> at your repo root for a single project.
Which bug does the agent read?
The one you name. Give its key, such as <code>WEB-142</code>, or paste a BugMojo share link, and Cursor uses the BugMojo tools to read that bug's replay summary, console and network logs.
Related articles
- Install BugMojo in Claude CodeInstall the plugin for nine skills, four subagents and three hooks, or add the hosted MCP server alone with one command. Either way, you sign in once in the browser.
- MCP tool referenceKnow exactly which tools a connected AI agent can call over MCP to read a BugMojo bug and act on it.
- Agent access, tokens, and the security modelUnderstand how an AI agent's access is scoped, what it inherits from you, and the guardrails that keep it in bounds.
- How AI agents work with BugMojoUnderstand the full loop, from a captured bug to an agent-proposed fix that a human verifies and closes.
- Install BugMojo in GitHub CopilotOne command in VS Code, plus the API-key path the cloud coding agent needs, because it cannot do a browser login.

