Privacy — ContextForge

What ContextForge stores

ContextForge is infrastructure that developers integrate into their own AI agents. When a developer calls the Remember API, ContextForge stores durable memories about that developer's end users — things like stated preferences, facts, tasks, and decisions — along with metadata (timestamps, importance/confidence scores, and a vector embedding used for semantic search). We also store the original raw content that was analyzed, so developers can audit exactly why a memory was created.

Multi-tenancy & isolation

Every memory belongs to exactly one project, which belongs to exactly one organization. All storage and retrieval queries are scoped by project ID at the database layer, and every API request is authenticated with a project-scoped API key — one customer's data is never visible to another.

Deletion

Developers (and, by extension, their end users) can delete individual memories at any time via DELETE /api/v1/memories/{id}, or delete every memory ContextForge holds for a given end user via DELETE /api/v1/memories with a user_id. Deletion is immediate and permanent — we do not retain deleted memory content.

Third-party processing

Memory extraction, conflict detection, and ranking are performed by Google's Gemini API. Content sent to Gemini is used only to produce the structured analysis ContextForge stores; ContextForge does not use customer content to train its own models.

Security

API keys are stored as salted one-way hashes, never in plaintext, and are shown to you only once at creation time. All traffic is served over HTTPS. See the README in our repository for a full breakdown of our security practices.

Questions

This is an MVP built for a hackathon/competition submission — reach out to the project maintainer with any questions about data handling.