Fulcra is a user-owned context backend for AI agents. Its core goal is to provide users with a unified, controllable context layer that brings together real-world data, recorded knowledge, files, and agent work, so that every agent can continue from where the previous agent left off.
I. Purpose Fulcra mainly solves the problem of fragmented context between agents. According to the official description, most agents are like "smart strangers": they do not know what has changed since the last conversation, and they cannot continue working based on what another agent has learned. When users switch between different applications, different agents, or different model providers, each system keeps only its own separate record, causing relationships and context to start over. Fulcra brings this scattered information together into a user-owned context, allowing agents to read from and write to the same context.
II. Main Capabilities
- Unified context: Brings real-world data, recorded knowledge, files, and agent work together into a user-controlled context.
- Context lake: At its foundation is a data platform that accumulates structured, semi-structured, and unstructured data for each owner; the data can come from multiple sources, agents, and applications, and accumulates over time.
- Cross-agent sharing: Every agent can contribute what it has learned, while all agents draw information from the same complete context; when switching agents or models, the context can migrate with the user.
- Viewable and controllable: The context is a browsable, human-readable view that shows what agents know about the user, including uploaded files, data learned by agents, and insights extracted from activity.
- Change log: Every update made by an agent is recorded, and users can track what happened and correct and control their own context.
- Agent memory, actions, and loops: Captures what agents remember, the operations they perform, and the repetitive workflows that turn context into follow-ups.
- Connectors and device data: Any connector can be added to bring application, device, and tool data into a shared context layer, such as events, notes, location, health, experiences, experiments, and more.
- Example use cases: Can connect finance, health, productivity, or custom agents.
III. How to Use The official text mentions "now recruiting: a small batch of builders to study loop engineering together with the Fulcra team." Regarding specific installation steps, API integration methods, or operating procedures, the materials do not currently provide them.
IV. Data Privacy and Rights Fulcra emphasizes that privacy and ownership are infrastructure, not an afterthought.
- End-to-end encConfidentiality: Data is encrypted both at rest and in transit, using industry-standard AES-256 encryption, so that even in the event of a breach, the data cannot be read without the key.
- Never sell data: The business model is based on subscriptions, not data brokerage; personal data will not be sold, licensed, or monetized.
- Access control: Provides fine-grained sharing controls, allowing you to decide who sees what, for example sharing sleep data with a doctor without exposing location, or sharing a project folder with collaborators while keeping other content private.
- User rights: Includes the right to export, allowing you to download context at any time in standard open formats, including connected or recorded content as well as superimposed agent work, with no proprietary lock-in, no waiting period, and no incomplete exports; the right to deletion, allowing you to request complete deletion of data; and the right to transparency, allowing you to understand what data has been collected, how it is stored, who has access to it, and how it is used.
V. Applicable scenarios Suitable for scenarios involving the use of multiple AI agents where they are expected to share user context and continuously accumulate it, such as finance, health, productivity, and other areas requiring continuity across tools and models. It is also suitable for users who want to control their own data, audit the sources of agent knowledge, and revoke access at any time.
VI. Limitations Regarding the specific supported connector types, platform compatibility, fee standards, pricing, available regions, technical integration details, etc., the materials do not currently provide this information.
