Guide
AI agents (MCP): connect Claude, ChatGPT, Claude Code or Cursor
jentry runs a remote MCP server, so an AI agent can work with your errors directly: "what broke since my deploy?", "show me the stack trace", "mark it resolved", "set up error tracking for this new app". The agent acts as you, with your role in each organization.
The server URL
https://jentry.app/mcpStreamable HTTP. Clients that support OAuth (claude.ai, ChatGPT, Claude Code, Cursor) only need the URL: they open jentry, you sign in and click Allow. Nothing to register anywhere.
Claude (claude.ai and desktop)
Settings → Connectors → Add custom connector, paste the URL, click Connect, sign in to jentry and Allow.
ChatGPT
Settings → Apps & Connectors → enable Developer mode (Advanced) → Create. Paste the URL, choose OAuth, sign in to jentry and Allow.
Claude Code
claude mcp add --transport http jentry https://jentry.app/mcp
# then, inside Claude Code: /mcp → jentry → authenticateCursor and other clients
{
"mcpServers": {
"jentry": { "url": "https://jentry.app/mcp" }
}
}A client that can only send a header can use a token from Settings → AI agents instead: Authorization: Bearer jmcp_…. Tokens are shown once; you can have up to 20.
What the agent can do
| Tool | What it does |
|---|---|
| list_projects | Your organizations (with your role) and their projects, with open issues and 24h events |
| get_project_setup / create_project | A project's DSN and the install + init code per platform; create a project for a new app (plan limits apply) |
| list_issues | Filter by status, project, environment, release, first/last seen and text; sort by recency or frequency |
| get_issue / list_issue_events | Stack trace with in-app frames and source lines, breadcrumbs, tags, request, release, environment — personal data trimmed |
| issues_since | What broke since a release (its last deploy) or a time: new issues, regressions, busiest issues, error volume vs before |
| resolve_issue / ignore_issue / reopen_issue / assign_issue / comment_issue | Triage, recorded in the issue's activity under your name |
| list_releases / get_error_stats | Releases with deploys; errors per hour or per day per project |
| list_alert_rules / set_alert_rule_enabled | Read alert rules; owners and admins can turn them on or off (turning off needs an explicit confirm) |
Security
- The agent only sees the organizations you belong to, with your role — removing you from an org cuts its access at once.
- Agent tokens are separate from organization API tokens and from DSN keys, and only their hashes are stored.
- OAuth uses PKCE; access tokens last an hour and refresh tokens rotate. Disconnect any app in Settings → AI agents.
- Resetting your password revokes every agent token and connection.
- Event data sent to the agent has emails masked and IPs, cookies, auth headers and request bodies removed.
release and environment in your SDK init, so the agent can answer "what broke since this deploy?" precisely.Frequently asked questions
Does jentry have an MCP server?
Yes. Point any MCP client at https://jentry.app/mcp; it signs in with OAuth or a token from Settings → AI agents.
Can an AI agent set up error tracking for a new app?
Yes. create_project makes the project and returns its DSN plus the install and init code for the app's platform; the agent can then verify the first event arrived with list_issues.
Try jentry free
Hosted error tracking & performance monitoring. Works with your Sentry SDKs — send your first event in minutes.