Dimension scores are derived from public data and fields; weighted into the composite. Reference only.
cerebral.work is an AI coding agent infrastructure project that is still in “coming soon” status. Its core idea is “context before the prompt”: before a prompt actually runs, a memory layer decides what context the agent should see. The currently available public materials are centered on three components in one repo: reveried, reverie, and cortex.
reveried is a long-running daemon that holds the memory store, exposes services over local HTTP on :7437, and also runs offline consolidation, referred to as the dream cycle. reverie is the memory layer managed by the daemon, with an emphasis on placement-aware storage and context integration. cortex is the CLI used by operators; it drives reveried over HTTP and manages local agent sessions based on tmux and git worktrees. Overall, it is more like a local context/memory middleware layer for AI coding agents than a chat or programming product aimed directly at end users.
The installation path is fairly engineering-oriented: it requires git and the Rust toolchain. The script builds the workspace and installs reveried, cortex, reverie-bench, and reverie-tracee into ~/.local/bin. Redis is recommended but not required; if unavailable, it falls back to the filesystem. A systemd user bus is recommended. Windows requires WSL2. In terms of interfaces, the documentation clearly mentions /health and cortex health --json for external monitoring, but does not disclose the full API. Pricing has not been made public. The project is labeled as proprietary, and commercial licensing, embedded use, and evaluation access all require contacting the team by email.
Its main strength is a very focused positioning: it attributes AI coding agent failures to context management and provides a clear three-layer split between daemon, memory layer, and CLI. The local installation does not use sudo and does not write outside $HOME, which is relatively restrained from an engineering perspective. The downside is that information is currently limited: there is no model list, performance benchmark, privacy policy, public pricing, or documentation on Chinese language support. The version is v0.9.13, and commands and parameters may change before v1.0, so production stability still needs to be validated.
It is better suited to engineering teams, researchers, and tool developers who are building or evaluating AI coding agent infrastructure, rather than ordinary programming users. Access from China cannot be determined from the available content. Installation depends on GitHub, the Rust ecosystem, and possibly Ollama/model environments, so the actual experience may be affected by network conditions. If you need a more mature alternative, you can compare it with built-in context mechanisms in LangGraph, MemGPT/Letta, Cursor, or Claude Code.
⚠ This review is compiled from public sources and does not constitute a purchase recommendation. Verify all facts on the vendor's official site. Verify on cerebral.work official site.
cerebral.work is an Unknown Site Builders provider. TG4G tracks its product information, an overall rating of 6.0/10, and a China-accessibility score of Workable. Click "Visit Official Site" to reach cerebral.work directly.