AI Agent (MCP)
Let Cursor, VS Code Copilot, or any MCP client read captured issues and resolve them.
The AI Agent connects your AI coding assistant to SnagRelay over the Model Context Protocol (MCP), so it can pull full bug context — screenshot, environment, console/network data — into a chat, and mark issues resolved once the reporter has tested the fix.
Setting it up
Bind the workspace
Open the project MCP tab and copy .cursor/mcp.json or .vscode/mcp.json. Commit
that file — it has the project URL only, no token.
If you use the VS Code / Cursor extension, Sign in as owner (or a teammate’s Authenticate VS Code / Cursor) writes this config when the folder doesn’t have it yet.
Sign in
Owners: Sign in as owner in the editor. Teammates: Authenticate VS Code / Cursor from Your access after a Team invite.
Enable the snagrelay server in Cursor Settings → MCP (or the VS Code MCP panel).
For CI, create a token on Account → API tokens — see API Keys & Tokens.
Use it from chat
Once connected, paste a tracker key (PROJ-123, #88, AB#456) or ask the assistant
to load a captured issue. It should call get_issue_context first when the ticket
came from SnagRelay.
Available MCP tools
| Tool | Purpose |
|---|---|
get_issue_context | Screenshot, console, HTTP summaries — SnagRelay-captured tickets |
find_issues | List captured issues when the user asked to list them (no key) |
mark_issue_resolved | Close the ticket after the reporter tests |
get_related_history | Past fixes for the same selector / endpoint / group |
The VS Code / Cursor extension sidebar is live
tracker columns, not find_issues.
Revoking access
- Team → Revoke on a person — invalidates their editor grant.
- Account → API tokens — revoke a CI / owner PAT.
- Changing
mcp.jsondoes not revoke anyone; that file has no secret.