AI audience
Agents need data logistics, not uncontrolled source access.
WWKG gives built-in agents, external agents, and worker-node agents a governed way to act in context. An agent can be bound to a workspace, use case, persona, and story; propose changes on a staging branch; emit task events into the workspace activity journal; trigger validation, review, ETL, or other agents; and merge only through the workflow that owns production knowledge.
Use case context tells the agent what situation it is in and which concepts matter.
Stories define permitted units of work and enforcement points for policies, regulations, workflow preconditions, and postcondition checks.
Agents and sub-agents can create their own what-if branches, work out alternatives, compare results, and promote the best option.
Task events and provenance make agent actions attributable across delegation chains, branches, commits, and produced facts.
Workspace discovery lets agents find useful knowledge products without scraping every source.
Bring your own agent or use ours
External agents can connect through tools, plugins, skills, MCP services, or SDKs. WWKG also exposes a site agent that can do what a user can do, including writing queries and creating workspaces. Worker nodes can grow into agentic execution nodes for use-case scoped tasks.
Agent work is branch work
The safe write pattern is the same for humans, pipelines, and agents: branch first, commit proposed graph changes, compare alternatives, validate, review, and merge. Every agent or sub-agent can create a branch for a what-if scenario, work it out independently, then let review agents, ETL steps, humans, or workflow gates decide what reaches production.
Agent actions are journaled
The workspace activity journal records task lifecycle events, delegation, failures, commits, and derived facts. That gives reviewers a queryable provenance chain for what an agent attempted, who authorized it, where it ran, and what it produced.
Context is not just retrieval
Retrieved text is not enough for operational agents. They need personas, outcomes, domain vocabulary, permissions, evidence, and rules. WWKG makes that context graph-native.
Continue reading