
Plexus
A typed knowledge graph that inverts the usual wiki model: your AI agent writes, you read. Instead of dumping markdown into folders, the agent creates structured entities (concepts, decisions, tasks, facts) and connects them with named, temporal relationships via MCP tools like save_entity, link_entities, and get_related. Think of it as persistent memory across sessions and agents. One conversation ends with Claude Code writing a handoff fact, the next morning Claude web loads the same graph and picks up where you left. Runs on-prem in Docker Compose with SurrealDB storage. Built for teams that want decision archaeology and cross-agent continuity without wrestling Notion or grep-ing Slack. The dashboard shows you the graph but has no edit buttons by design.
- tools
- 15
- transport
- HTTP
- last updated
- Apr 27, 2026
Tools
15 toolsPublic tool metadata for what this MCP can expose to an agent.
save_entityCreate a new entity
Create a new entity
No parameter schema in public metadata yet.
get_entityFetch a single entity by id
Fetch a single entity by id
No parameter schema in public metadata yet.
list_entitiesList with filters (kind, context, status)
List with filters (kind, context, status)
No parameter schema in public metadata yet.
search_entitiesBM25-ranked full-text search with optional highlighting
BM25-ranked full-text search with optional highlighting
No parameter schema in public metadata yet.
update_entityUpdate with optimistic locking (`expected_version`)
Update with optimistic locking (`expected_version`)
No parameter schema in public metadata yet.
archive_entitySoft-delete (status → archived)
Soft-delete (status → archived)
No parameter schema in public metadata yet.
link_entitiesCreate a temporal edge between two entities
Create a temporal edge between two entities
No parameter schema in public metadata yet.
unlink_entityInvalidate an edge (`valid_to = now`)
Invalidate an edge (`valid_to = now`)
No parameter schema in public metadata yet.
get_relatedTraverse edges with direction + point-in-time filter
Traverse edges with direction + point-in-time filter
No parameter schema in public metadata yet.
list_kindsQuery the kind registry
Query the kind registry
No parameter schema in public metadata yet.
list_relationsQuery the relation registry
Query the relation registry
No parameter schema in public metadata yet.
context_loadSession warm-up — 20 most recent entities + registries
Session warm-up — 20 most recent entities + registries
No parameter schema in public metadata yet.
lint_graphOrphans + duplicate-title check
Orphans + duplicate-title check
No parameter schema in public metadata yet.
list_skillsCompact index of all skills
Compact index of all skills
No parameter schema in public metadata yet.
load_skillLoad one skill by name, trigger phrase, or BM25 fallback
Load one skill by name, trigger phrase, or BM25 fallback
No parameter schema in public metadata yet.
Sponsors
Readme
plexus (_graph)
A typed, on-prem knowledge graph for AI agents. Read-only for humans, write-only for agents, spoken over the Model Context Protocol.
Part of the pcas.io product line.
plexus graph view — Apollo-program demo graph
TL;DR
Knowledge management has been a document sport for twenty years — folders, notes, wikis, editor wars. plexus inverts that.
In plexus, every piece of information is a typed entity — concept, decision, fact, project, task, document, skill, … — with hard, temporally-valid edges between them. No free-form text, no forgotten tags, no "I'll clean that up later." Writing and linking happen exclusively through MCP — from your agent. The dashboard shows you the graph; it has no edit buttons.
That gets you three things:
Details
- Repository
- github.com/pcas-io/plexus
- Category
- Documents & Knowledge
- Transport
- HTTP
- Registry
- MCP registry, active