What vivac MCP server does
The vivac MCP server is presented as a provenance system for coding work. It represents goals, tasks, decisions, findings, and session checkpoints as nodes in a tree, with parent-child relationships showing how later work originated. This structure is intended to answer both where a project is and why the current work exists.
The supplied material describes vivac primarily through its command-line interface. brief shows the active project and lane, the current position in the tree, standing decisions, and the most recent checkpoint. why walks upward from a node to explain its lineage. done checks whether required child findings remain open before allowing closure, while --force records an explicit override. Decisions can retain rejected alternatives, superseded decisions, and the principle used to judge them.
How it works
A session-start hook runs vivac session start, giving an agent an initial orientation. The resulting brief is deliberately compact and includes a token count, tree depth, and parked-item count. A project can therefore resume with a focused representation of its current state rather than reconstructing context from chronologically scattered notes.
The model also supports abandoning a premise. Descendants are marked as affected, but rescued items remain attached to their original parent so their history is not rewritten. The documented examples show work spanning goals, tasks, decisions, and findings, including open and closed states.
Setup and configuration
The provided material identifies vivac as a Rust project distributed through crates.io and shows a minimum supported Rust version of 1.89. It includes the vivac web command, which opens a browser view of projects on the machine. That view can show project activity, changes since the last visit, node lineage, and the complete tree.
No MCP installation command, server launch command, configuration file, transport, or environment variables are specified in the supplied material. The vivac MCP server should therefore not be assumed to work with a particular desktop client without additional documentation.
Tools and capabilities
Documented commands and functions include:
vivac brief for current project orientation.
vivac why for tracing a node back through its parents.
vivac done for validating closure conditions.
vivac abandon for invalidating a premise and handling its descendants.
vivac web for browsing project trees locally.
- Session-start capture for automatically placing project state into the agent’s context.
Limitations and notes
The vivac MCP server material does not claim that all relevant facts will be captured automatically. Agents still decide what to save, what counts as a finding, and when to consult the tree. vivac is described as reducing the context needed to resume work, not as eliminating file reading, reasoning, or explanation.
The excerpt also does not document MCP-specific tools, supported clients, network access, authentication, or a server protocol. Those details require confirmation from the repository’s MCP documentation before integrating vivac into an MCP client.