Talorys Deploys Private Autonomous AI Agents on Cloudflare Free Tier
Talorys is an open-source personal AI assistant built to deploy inside a user's Cloudflare account via Cloudflare Workers and Durable Objects. It manages persistent SQLite memory, background routines, and streaming chat without external databases or server costs.
Why it matters
Developers can spin up a production-grade, zero-cost private agent with durable storage and background routines using a single command.
TL;DR
- 01Talorys deploys a self-contained AI assistant directly into Cloudflare's free tier with zero external database dependencies.
- 02Persistent memory and scheduled automations run inside an SQLite Durable Object using native alarms.
- 03Core organizational features remain accessible even when daily LLM inference quotas are fully consumed.
Key facts
- Chat Model
- @cf/zai-org/glm-4.7-flash
- Stack Components
- Cloudflare Pages, Hono Worker, Durable Objects (SQLite)
- Minimum Node Version
- 20.18+
- Pricing Model
- Cloudflare Workers Free tier compatible
Serverless Architecture on Cloudflare Agents SDK
Talorys packages an entire personal assistant into a single Cloudflare account without spinning up external servers or databases like D1 or Vectorize. Everything is coordinated inside a TalorysAgent Durable Object that pairs embedded SQLite storage with Cloudflare Workers AI using @cf/zai-org/glm-4.7-flash. The frontend, built with React and Vite, runs on Cloudflare Pages, routing traffic through /api/* via a private service binding with workers_dev: false.
Durable Memory and Zero-Cost Routines
State persistence relies on transactional SQLite migrations inside the Durable Object. Memory is selective: only relevant preferences and facts are injected into active prompts to conserve context limits. Routine automations and recurring reminders run on native Durable Object alarms rather than external cron triggers, guaranteeing execution without keeping processes awake or invoking the LLM for plain task lists.
Deployment and Guardrails
Installation executes via an interactive or scriptable CLI (npx create-talorys@latest), validating Node.js 20.18+, configuring PBKDF2-SHA256 password authentication, and verifying storage health before any inference runs. Built-in guardrails cap output tokens, context summaries, and daily AI invocations to prevent exceeding Cloudflare's free daily Neuron allowances.
Try it in 2 minutes
npx create-talorys@latestbash
✓ When to use
- Deploying a zero-maintenance private personal agent without managing physical servers or a VPS.
- Managing personal tasks, durable memories, and scheduled alerts on free cloud infrastructure.
- Experimenting with agentic Durable Object architectures using Cloudflare Agents SDK.
✕ When NOT to use
- Multi-user or enterprise collaboration environments requiring complex team roles and access tiers.
- Workloads demanding complete hardware isolation decoupled from proprietary Cloudflare APIs.
- High-throughput inference tasks that frequently exceed standard daily Workers AI free tier quotas.
What to do today
- Run npx create-talorys@latest to bootstrap the personal agent stack in your Cloudflare account.
- Configure token and reasoning caps in Settings -> AI to optimize usage against free Neuron limits.
- Download talorys-backup.json periodically under Settings -> Privacy to maintain local state backups.
What the community says
“Did a side project on Workers and the self host test for me was whether I could run it off Cloudflare at all. Durable Objects and KV bindings meant no, so it was really just deploy it yourself.”
“To me, 'self-hosted' means, either I use a rented hardware or I use a purchased hardware. If the software can be switched to a rented or purchased hardware... then it is 'self-host-able' IMO.”