Cloudflare Workers AI Free Tier Can Run Out in 35 Messages. I Did the Math.

Description and use

Talorys, a personal agent on Cloudflare Workers AI, can exhaust the free 10,000‑neuron allowance in 35 full‑context messages, limiting long conversations. Adjusting token limits and request caps keeps usage within the free tier.

Problem

Talorys, a personal agent built on Cloudflare Workers AI, can exhaust the free 10,000‑neuron daily allowance in as few as 35 full‑context messages, limiting its usability for extended conversations.

Outcome

Talorys demonstrates that a fully serverless personal assistant can run on Cloudflare’s free Workers AI tier, but the neuron budget limits long conversations to about 35 full‑context messages per day. Adjusting token limits and request caps can keep usage within the free allowance, while the free tier’s CPU and request limits are not reached.

Workflow

• Clone the Talorys repository from GitHub.
• Run npm ci, npm run build, and npm run dev to start a local development environment using Wrangler and Vite.
• Interact with the agent via the /api/chat endpoint to create tasks, notes, etc., observing the number of AI requests and token usage reported by /api/usage.
• Calculate neuron consumption per message using the model’s input and output neuron rates (5,500 neurons/1M input tokens, 36,400 neurons/1M output tokens).
• Adjust configuration values in packages/shared/src/constants.ts (e.g., maxOutputTokens, maxContextTokens, maxAiRequestsPerDay) to reduce neuron usage.
• Monitor the actual neuron count in the Cloudflare dashboard under Workers AI and compare it to the in‑app usage estimate.
• Export a backup (talorys-backup.json) from Settings → Privacy to mitigate lock‑in.

Evidence

With Talorys's default settings, a long conversation can use up Cloudflare's free AI allowance in about 35 messages.

Source

Talorys is an MIT‑licensed personal agent that showed up on HN this week, and it's built to live inside that allowance: chat, memory, tasks, notes and reminders, with no VPS, subscription or paid database behind it.

Source

The defaults in packages/shared/src/constants.ts are maxContextTokens: 6000, maxOutputTokens: 1024, maxAgentIterations: 4, maxAiRequestsPerDay: 200.

Source

Short chat: 2,000 in / 200 out ~18 neurons, ~550 calls per free day; Full context: 6,000 in / 1,024 out ~70 neurons, ~140 calls; Full context, 4 agent iterations ~280 neurons per message, ~35 messages.

Source