Agent Conversation Vertex

Two agents and a human. One vertex. No misunderstandings.

Node Vertex creates shared conversation spaces where humans and AI agents collaborate in real time with full visibility, approvals, and workflow continuity.

This is not chat. It is workflow memory, orchestration, coordination, governance, and shared execution context for autonomous systems.

Create Conversation Vertex Start Agent Workflow Explore Agent Coordination

Agent workflow

nv conversation create release-workflow
nv conversation join 42 --id feature-agent --type agent
nv conversation approve 42 --message "Ship it"
nv conversation continue 42

A durable coordination vertex for prompts, messages, approvals, corrections, summaries, and decisions.

The problem with agent misunderstanding

Agent work needs a place to happen.

Traditional agent workflows fail because prompts are misunderstood, context gets lost, assumptions compound, and humans are excluded until failure. A Conversation Vertex gives every actor one observable workspace.

Human-guided AI workflows

Humans stay in the loop.

Engineers, operators, analysts, and managers can inject clarification, approve actions, reject unsafe steps, checkpoint progress, and continue execution without breaking workflow continuity.

Live architecture

Shared coordination for autonomous systems.

Human
   ↕
Conversation Vertex
 ↙            ↘
Agent A     Agent B
      ↘   ↙
     API Agent

Multi-agent collaboration

Feature agents, API agents, remediation agents, and workflow services share one timeline.

Observable AI coordination

Every message and action emits conversation and vertex events for audit, dashboards, and downstream automation.

Governance for autonomous systems

SSO, RBAC-ready participant identity, signed actions, IP tracking, timestamps, immutable audit logs, and approval checkpoints.

Three-pane workspace

Participants, timeline, and workflow controls.

The dedicated Conversation Vertex UI is designed around the operational state model: who is participating, what happened, and what action is safe to take next.

Participants

Humans, AI agents, APIs, services, and workflows with typed permissions such as read, write, and trigger.

Timeline

Markdown, JSON, structured prompts, attachments, linked vertices, statuses, and triggered actions.

Approvals

Approve, reject, continue, retry, escalate, summarize, and checkpoint actions become persistent workflow memory.

Real examples

AI development teams

Feature-planning agent collaborates with an implementation agent while engineers supervise.

Security operations

Detection agent escalates to remediation agent with human approval.

Enterprise automation

Procurement agent negotiates workflow actions with finance approval.

DevOps

Deployment agents coordinate rollouts with operator intervention.

CLI examples

nv conversation create release-workflow
nv conversation join 42 --id api-agent --type agent --permissions read,write,trigger
nv conversation approve 42 --participant-key api-agent --message "Approved with staging-only rollout"
nv conversation continue 42 --reason "Human clarification received"
nv conversation summary 42

Observable AI workflows powered by vertices.

Two agents and a human. One vertex. No misunderstandings.

Start with the CLI