Your agent
reads the report

The gap: your agent can't see what your user saw

A coding agent is good at fixing a bug it can read. A one-line ticket gives it nothing to read. The MCP server gives the agent the same evidence a human debugger would open.

What your agent can call

• get_issue_context: everything about a bug, by key
• find_issues: search and list open bugs for the project
• get_issue_payload: one full request or response body, on demand
• get_related_history: similar past bugs and the changes that fixed them
• mark_issue_resolved: close the bug and tell the person who reported it
• report_progress: post a status update while the agent works
• whoami: check which account and project the agent is connected to

Set it up

Add the server once, using the command for your agent:

• Cursor: /add-plugin snagrelay
• Claude Code: claude mcp add --transport http snagrelay https://app.snagrelay.com/mcp
• Codex: codex mcp add snagrelay --url https://app.snagrelay.com/mcp

The fix, then resolve loop

1. A user reports a bug; the ticket gets a key such as SHOP-142.
2. You tell your agent: fix SHOP-142.
3. The agent calls get_issue_context, reads the failing request and error stack, and edits the code.
4. The agent calls mark_issue_resolved.

When the agent resolves the bug, the person who reported it gets a "fixed" email, so nobody has to follow up by hand. Reports are grouped, and get_related_history shows the agent earlier bugs like this one and what fixed them: related reports and history.

MCP is included on every plan, including Free. See SnagRelay for coding agents for the full flow and the install snippet for each agent.