Build-time guide for coding agents
Wiring JamBase Data into your own codebase with help from Cursor, Claude Code, Windsurf, or Replit Agent? Paste one prompt and your agent scaffolds the env, auth, Hello World, paginated bulk seed, and dateModifiedFrom delta sync against your database of choice.
The single file your agent reads is /agents.md. Hand it to the agent or paste the starter prompt below.
Raw file URL: https://data.jambase.com/agents.md
If you want an AI assistant to answer music questions inside ChatGPT, Claude, Cursor's chat, Gemini, or Replit — "who is playing at Red Rocks this August?" — you want the MCP server at /mcp instead. That's the runtime surface. This page is for building software on top of JamBase Data with a coding agent.
Copy-paste into Cursor, Claude Code, Windsurf, or Replit Agent. The agent self-discovers JamBase Data via agents.md + llms.txt and scaffolds env, auth, Hello World, bulk seed, and the dateModifiedFrom delta loop on the first run.
You are scaffolding a JamBase Data integration into this codebase. 1. Fetch https://data.jambase.com/agents.md and read it end-to-end before writing any code. It is the build-time counterpart to llms.txt and contains the env-var convention, auth shape, pagination etiquette, suggested local schema, bulk-seed loop, and dateModifiedFrom delta-sync loop. 2. Fetch https://data.jambase.com/llms.txt as the declarative reference for endpoints, plans, and quotas. Then, in this codebase: - Add JBD_API_KEY and JBD_BASE_URL (default https://api.data.jambase.com/v3) to .env and .env.example, and to .gitignore if missing. My key starts with "jbd_" — I will paste the value into .env myself. - Implement the tiny fetch wrapper from §4 (Bearer auth, 429 + Retry-After, 5xx exponential backoff with jitter). - Implement the Hello World from §3 against /v3/events?geoMetroId=jambase:1 and run it once so I can see real rows. - Create the suggested Postgres schema from §7 (events, artists, venues, event_performers, event_tombstones, sync_state) using **[MY MIGRATION TOOL — drizzle / prisma / knex / plain SQL]**. - Implement the bulk seed from §8 and the dateModifiedFrom delta loop from §9, scoped to **[MY GEO — e.g. geoMetroId=jambase:1 for NYC]**. - Wire the tombstone sweeper from §10 as a nightly job. - Stop and ask me before adding any extra dependency, framework, or background-job runner the codebase does not already use. When you are done, summarize what shipped, which sections of agents.md you applied, and what is still open (e.g. tombstone sweep cadence, attribution UI per §16).
Replace the bracketed slots with your details, then drop the prompt into your agent of choice.
Seventeen sections, plan-tier-annotated. Your agent reads top-to-bottom and produces a working ingest pipeline.
v1 is the single canonical agents.md file plus this landing page. A forkable starter repo with a "Deploy to Replit" button is a planned v1.1; we shipped the file first to measure pull-through before committing to a maintained starter.
If you'd find a forkable starter useful right now, drop us a note at /contact — concrete asks bump it up the queue.