Getting Started with Telebugs MCP

Telebugs exposes a Model Context Protocol (MCP) server that lets AI coding tools read your error data and take action on your behalf — directly inside your editor.

Instead of copying stack traces into a coding agent, you can hand it safe Telebugs identifiers and let it retrieve the authorized context through MCP. Connected agents can:

• Inspect full error reports with backtraces and context
• Fetch one-call investigation context for a coding agent
• Search and filter error groups using the same powerful query syntax as the REST API
• Resolve, mute, assign, and annotate issues
• Add notes and manage your error workflow

The MCP integration mirrors the REST API: the same projects, groups, reports, and notes are available, scoped to your account and project memberships.

Quickstart

Get value in under two minutes:

1. Connect your editor Follow the step-by-step instructions in Connecting AI Tools for Cursor, Windsurf, or Claude.

2. Try a prompt Once connected, ask your AI something like:

   “List my Telebugs projects and show the open error groups in Production with the most reports this week.”

   Or:

   “Find the top 5 unresolved error groups in production and summarize what’s happening.”

   Or open an error group in Telebugs, select Copy for LLMs, and paste the identifier-only prompt into your coding agent. See Agent Investigations.

3. Take action You can then say:

   “Resolve group 42 with the note: ‘Fixed by PR #847 — deployed in v1.4.2’”

   Or ask it to assign an issue to a teammate, add context, or investigate a specific report.

What You Can Do

Connected AI tools can:

• List your projects and error groups
• Search and filter errors with the same query syntax as the REST API
• Fetch full error reports with backtraces, breadcrumbs, and request context
• Investigate a group or event with one read-only context call
• List project members to find the right assignee
• Resolve, mute, assign, and annotate error groups
• Create, list, and delete notes on error groups

MCP Endpoint

Your Telebugs instance serves MCP at:

https://your-telebugs-instance.com/mcp

Discovery

MCP clients can discover the server automatically:

curl https://your-telebugs-instance.com/.well-known/mcp.json

Example response:

{
  "mcp_endpoint": "https://your-telebugs-instance.com/mcp",
  "authorization_servers": ["https://your-telebugs-instance.com"],
  "protocol_version": "2025-06-18",
  "transport": "streamable-http"
}

Authentication

MCP supports two authentication methods:

1. OAuth 2.0 (recommended for AI tools like Cursor and Claude) — see Authentication
2. API key — use your existing REST API key as a Bearer token (useful for scripts)

For step-by-step setup in your editor, see Connecting AI Tools.

Managing Connected Apps

After authorizing an MCP client, you can review and revoke access from Account Settings → Connected apps.

Available Tools

Telebugs MCP tools are grouped by resource:

Security Notes

• Access is limited to projects you belong to.
• OAuth uses telebugs.read and telebugs.write scopes.
• Application-supplied error data is marked as untrusted so AI tools treat it as debugging evidence, not instructions. See Reports for details.

Next Steps

• Read the Authentication guide to understand OAuth vs API keys
• Follow Connecting AI Tools for your specific editor
• Explore the available tools in the Error Groups reference
• Hand an error to a coding agent with Agent Investigations