Fulcra

Fulcra is a user-owned context backend that lets multiple AI agents share memory, data, files and work output in one auditable repository. Access is via MCP, CLI, API and Python, with a free 5 GB start.

Fulcra is a context backend for people who run more than one AI agent and are tired of each one knowing only its own fragment of the picture. Instead of letting every chat app, model provider or specialized agent keep a private record of its user, Fulcra stores real-world data, recorded knowledge, files and the output of agent work in a single repository that the user owns. The accumulated material is described as a context lake: structured, semi-structured and unstructured data gathered from many sources, agents and applications over time. In practice, agents connect to that repository and both read from it and write back to it. A signal such as a campaign metric, a browsing-history item or a sensor alert lands in Fulcra, and any connected agent can use it without needing direct access to the original source. The homepage demo shows one agent turning an article and several signals into a brief, handing it to a work agent that builds the deliverable, and then storing the result as durable context. Access is exposed through MCP, a CLI, an API and Python, and a Claude connection page is listed. Context Mobile collects selected data from a phone, while Context Web offers a desktop view for inspecting and controlling what is stored. Ownership and visibility are the central pitch. The context is presented as a browsable, human-readable view, every agent change is logged, and access can be granted or revoked at any time, so switching models or tools does not mean starting over. Data is described as encrypted at rest and in transit with AES-256. The company says it is built by people behind HUMAN and Veracode. Among alternatives, Fulcra sits closer to infrastructure than to the built-in memory features of individual assistants. It suits developers, agent builders and power users who coordinate several agents across platforms and want portable, auditable memory. A free tier of 5 GB is offered, and an early-access cohort for builders is also being enrolled.