Caide

Local Storage & Data Privacy

Learn how Caide stores workspace state in local SQLite databases with zero cloud tracking.

Last updated 2026-10-29

Local Storage & Data Privacy

Caide is fundamentally engineered as a local-first application. Unlike cloud-hosted AI coding assistants that upload your entire workspace to proprietary servers, Caide stores all state strictly on your local disk.


What Lives on Your Machine

All application state is stored locally inside your OS user data directory (~/.config/caide or ~/.caide):

Data TypeStorage EnginePurpose
Threads & TurnsSQLite via Drizzle ORMConversation history, user prompts, assistant turns, token counts
Goal TreesSQLite via Drizzle ORMHigh-level goal tracking, sub-tasks, and execution checklists
Git CheckpointsEmbedded Git worktreePoint-in-time snapshots for 1-click rollback after any turn
API Keys & SecretsOS safeStorage / KeychainAES-256 encrypted provider keys and tokens
Project ConfigurationsJSON in .caide/Project-level framework settings, dev-server ports, and rules

Context Compaction & Memory

As chat conversations grow, Caide employs an intelligent context compaction algorithm:

  • When a thread approaches the model's context window limit, Caide identifies completed turns and summarizes intermediate reasoning steps into dense, structured checkpoints.
  • Critical architectural decisions, file paths, and current goals are preserved.
  • Unnecessary verbose tool outputs (e.g. 500 lines of build logs) are compacted into a single status confirmation, saving tokens and maintaining model reasoning sharpness.

Zero Telemetry Guarantee

  • Caide contains zero third-party analytics trackers (no PostHog, Mixpanel, or Google Analytics).
  • Your source code is never used to train models.
  • Network requests leave your machine only when explicitly making API calls to your configured model provider or when deploying to your chosen hosting service (Vercel / GitHub).
Ready to build with Caide?

Join our early access whitelist to test native desktop builds, copy-on-write database branches, and DeviceLab.

On this page