Engineer Ships Production AI Chatbot for $0 Across 11 Free Tiers
An AI engineer says he shipped Rune Score, a production sports analytics chatbot, for $0 by stacking free tiers from Figma, GitHub, Claude Code, Codex, Cloudflare, Railway, Clerk, Neon and more.
By Grace Kim
4 min read

What's News
- The engineer built the Rune Score sports analytics chatbot for $0 using eleven free tiers.
- GitHub's free plan includes 2,000 Actions minutes per month on standard runners for private repos.
- Clerk's free plan covers up to 50,000 monthly active users at no cost.
- Codex's $0 plan runs GPT-5.6 Terra, capped by usage rather than time.
- Claude Code's free chat plan excludes the coding assistant; a Guest Pass unlocks one week of Pro-level access.
An AI engineer says he built and shipped a production-grade sports analytics chatbot, Rune Score, for exactly $0 — by stacking free tiers from eleven vendors rather than paying for any layer.
The author, who is currently building ChatMakr, an AI-powered platform that creates multi-agent chatbots in seconds, laid out the full toolchain. From the outside, he says, Rune Score looks like a simple chatbot. Under the hood it's a distributed system handling sports data freshness, user rephrasing, extraction failures, LLM call costs, streaming, memory, fallbacks and security.
His three headline findings:
- Every layer of a modern chatbot, from the logo to the database to the voice output, now has a genuine free tier, not just a trial.
- Almost all of these tools cap usage rather than time — the limit you hit first is volume of users or requests, not a countdown clock.
- The one layer worth watching closely is the coding assistant itself: Claude Code's free chat plan doesn't include it, so sustained use eventually means either a Guest Pass or leaning more on Codex's own free plan.
What's actually in the stack?
Branding — Figma. He generated the chatbot's logo from a text prompt describing the style and mood he wanted, then iterated inside Figma instead of hiring a designer. Figma's free plan covers individual use with unlimited file storage.
Code hosting — GitHub. The project lives in a private GitHub repository — the same repo both Claude and Codex connect to. GitHub's free plan includes unlimited private repositories with a limited feature set and 2,000 minutes a month of Actions on standard runners for private repos. Standard runners on public repositories stay free regardless of usage.
Development — Claude Code, with Codex as judge. Claude Code built the backend and the AI agents. The engineer never let it touch the main branch: every task ran on its own branch, and nothing merged until he reviewed the diff and approved the pull request. The genuine free route into Claude Code is a Guest Pass — someone already paying for Claude Max can send one, unlocking a full week of free Pro-level access.
Codex got a different job on the same repo. It never wrote production code; it reviewed what Claude shipped, acted as an independent judge on pull requests, and drafted the next development prompt handed back to Claude — one model planning, another executing. Codex has a genuine $0-a-month plan running GPT-5.6 Terra, OpenAI's cost-efficient model for that tier, capped by usage rather than time.
What does the infrastructure cost?
Domain and security — Cloudflare. He bought the domain through Cloudflare, which sells registration at cost with no markup, and kept DNS in the same place. Cloudflare's free tier includes DNS, DDoS protection and basic security features, plus rate limiting rules to stop any single source from hammering the chatbot's endpoints.
Deployment — Railway. Backend and frontend both run on Railway, with a separate test environment. New accounts get a one-time free credit grant before any card is required, and Railway doesn't put services to sleep by default — no cold start tax on the free tier unless you enable the optional Serverless behavior yourself.
Authentication — Clerk. The free plan covers up to 50,000 monthly active users at no cost. He built the entire flow in Clerk's test environment first, trying edge cases without creating real accounts or sending real emails.
Database — Neon. Conversations, users and application data sit in Neon, a serverless Postgres that scales down to zero and stops costing anything when nobody's using it. The free tier includes enough storage and monthly compute hours for a real early product, with no credit card required.
What about observability, research and voice?
Observability — Langfuse. Langfuse traces which prompts fired, where agents failed and what they cost. The free Hobby plan includes a monthly allowance of tracing events enough to debug and monitor a real chatbot without paying.
Web research — Exa and Firecrawl. Both are wired into the agents' runtime, so the chatbot decides when to look something up. Exa handles search and gives new accounts starting credit plus a recurring monthly allowance with no payment method; it also ships a browser playground for testing queries. Firecrawl turns pages into clean text the agent can read, with a monthly credit allowance and no card needed.
Model testing — NVIDIA. Before committing to a model, he used NVIDIA's API Catalog, which gives any developer free credits to test a wide range of hosted models through its own endpoints — comparing options before locking in the production choice.
Voice — Azure AI Speech. The chatbot talks back via text-to-speech, switching between a female or male neural voice. The free tier includes a monthly allowance of neural voice characters enough for genuine daily conversations before any bill appears.
The catch, per the author, is narrow: usage caps will bite before any trial clock runs out, and the coding-assistant layer is the only one where a truly free sustained path doesn't exist — plan around Guest Passes or Codex's GPT-5.6 Terra tier.
Original: figma.com
More from Grace Kim
Show full bio
Market editor covering people, launches and day-to-day practice in entrepreneurship, startups and venture funding at Business Bearings.
2,447 articles