Add analytics to Kiro.
Kiro is AWS's agentic IDE, and it reads MCP servers from .kiro/settings/mcp.json in the workspace or ~/.kiro/settings/mcp.json for every workspace. Add smolanalytics there with your org token and Kiro can both instrument your app and answer questions about it: funnels, retention, paths, what broke today, computed from your own events, in the same window where you write the spec. Save the file and Kiro reconnects, no restart.
{
"mcpServers": {
"smolanalytics": {
"url": "https://smolanalytics.com/api/mcp",
"headers": { "Authorization": "Bearer YOUR_ORG_TOKEN" },
"disabled": false,
"autoApprove": ["overview", "funnel", "retention", "whats_notable"]
}
}
}Workspace config is .kiro/settings/mcp.json, user config is ~/.kiro/settings/mcp.json, and workspace wins where the two overlap. Kiro reconnects the changed server when you save. Self-hosting? Replace url and headers with command: the smolanalytics binary, args: ["mcp"].
how do I add analytics to kiro?
Kiro is AWS's agentic IDE, and it reads MCP servers from .kiro/settings/mcp.json in the workspace or ~/.kiro/settings/mcp.json for every workspace. Add smolanalytics there with your org token and Kiro can both instrument your app and answer questions about it: funnels, retention, paths, what broke today, computed from your own events, in the same window where you write the spec. Save the file and Kiro reconnects, no restart.
Kiro's loop is spec first: requirements, design and tasks under .kiro/specs, then the agent works the task list. Instrumentation belongs in that list. With the MCP server connected, ask Kiro to add tracking for the flow you just specced, then ask what happened after it ships: "which step of the signup funnel drops hardest this week?" runs the funnel tool and returns the computed report. A steering file in .kiro/steering that pins your event names (signup, activate, checkout) keeps every later spec using the same ones, which is the thing that quietly breaks analytics as a codebase grows.
autoApprove is safe here because the tools are deterministic reports, never generated SQL. overview, funnel, retention and whats_notable only read, and a CI agreement test asserts that the answer Kiro reads matches the dashboard and the HTTP API for the same question, every build. Worth knowing: smolanalytics connect auto-wires Cursor, Claude Code, VS Code, Windsurf and Cline, but not Kiro, so the block above is the install.
If the thing you are building in Kiro is itself an agent, the same connection covers it. agent_tools, agent_errors and agent_conversations give tool-call error rate, latency at p50/p90/p99, the error taxonomy, and where conversations get abandoned, all computed from your events rather than inferred. Open source (MIT), self-host free or $49/mo hosted for 1M events, and Kiro answers on the model you already pay for, so we never meter the AI.
Honest pricing: 14-day full trial, no credit card. Then Pro $49/mo, never metered on seats or sites. Overage is $8/million with an emailed receipt, the dashboard never locks, and self-hosting the single Go binary is free forever (MIT).
Add analytics to Kiro tonight.
One snippet, or one endpoint. Tomorrow morning the verdict tells you which part of the funnel to fix.