Built for coding agents
llms.txt, an MCP server, self-fixing errors, and a one-command self-test — so an agent can integrate the backend itself.
Triggair is designed so a coding agent can wire it end-to-end. Point your agent at the docs and it has everything it needs.
The published artifacts
/llms.txt is a concise index; /llms-full.txt is the full guide (recipes + these feature guides + error docs); /openapi.json is the machine-readable API definition; /dropin is a copy-paste prompt. All are keyless and edge-cached.
The MCP server
The Model Context Protocol server lets an agent configure games, boards, economy, flags, moderation, and more — and read analytics — directly, without dashboard round-trips. Its tools (triggair_create_game, triggair_configure_leaderboard, triggair_verify_integration, triggair_search_docs, …) mirror the dev API with validated arguments.
Connect the MCP server
It's a streamable-HTTP MCP endpoint at POST /v1/mcp — point any MCP client (Cursor, Claude, Windsurf, …) at it and authenticate with your developer session token (copy it from app.triggair.com). The token authorizes exactly as the dev API does: an agent only ever touches games you own.
// ~/.cursor/mcp.json (or your client's MCP config)
{
"mcpServers": {
"triggair": {
"url": "https://api.triggair.com/v1/mcp",
"headers": { "Authorization": "Bearer <your session token>" }
}
}
}Errors that fix themselves
Every failure is a TriggairError with { code, message, agentHint, requestId }. The agentHint tells you how to fix the call; rate_limited/5xx/network retry automatically. Each code links to /docs/errors/<code>.
Close the loop
After wiring, the MCP tool triggair_verify_integration (or the dashboard verify runner) live-probes the key services and reports pass/fail with fixes — so an agent knows it's actually done, not just that the code compiled.