NucovoChoose the teammates.
Start a project-specific direct room with one employee or a group room with two to six. The participants and coordinator are durable records.
Talk in a room.
Get accountable work.
Work directly with a named employee or bring a small agent group together. Every request enters the same task, evidence and independent-review system as the rest of your company, and routines can start that work on a schedule or after a recorded company event.
This collaboration layer is implemented inside Nucovo. It does not depend on a private Grok Bot API or claim xAI’s hosted computer and live-takeover features.
A message can start work.
It cannot skip accountability.
Give the room real work.
A prompt is saved with a ready task for a selected participant. A different active employee is assigned to review it against the current project contract.
Let the normal worker run.
The task uses the employee's versioned model, including xAI Grok when configured, together with its permissions, budget, skills and MCP tool grants.
Follow evidence, not theatre.
The room shows linked tasks, attempts, artifacts, reviews and approvals. Delegated child work remains attached to the conversation that caused it.
Direct and group roomsOne project.
One project.
The right people.
A direct room has exactly one teammate. A group has two to six named participants and one of them coordinates the first request. The room stays attached to its original company and project.
Follow-up prompts create further tasks in the same room. If an employee delegates child work, the handoff inherits the conversation link so the original request and resulting work remain traceable.
The interface reports real task and review state. It does not synthesize an instant reply before a model run has happened.
Repeat the useful work.
Keep every run inspectable.
Interval, daily and weekly schedules
Select the employee, reviewer, project prompt and time zone. Pause or resume a routine, run an enabled routine immediately, or limit how often it can create work in a day. After an outage, the scheduler advances from the present instead of flooding the queue with every missed occurrence.
Respond to recorded operating events
Trigger from a narrow allowlist covering project, task, artifact, review, approval and meeting events. Small exact-match filters select the cases that matter. The task receives a bounded source snapshot labelled as untrusted data, and lineage guards prevent a routine from triggering itself forever.
Ordinary tasks, every time
A routine creates a task with an owner, different reviewer and current contract pin. Model, skill, MCP, allowance, evidence and approval controls continue to apply. Scheduling is never a permission bypass.
Exact versions and idempotent triggers
Edits create immutable definition versions. Each trigger records its version, source event or scheduled time, outcome and resulting task. Unique trigger keys stop duplicate claims, while history identifies created and skipped runs with the reason for each skip.
xAI model accessUse Grok.
Use Grok.
Keep Nucovo in control.
Configure an employee with the first-party xAI provider, an installation XAI_API_KEY and a model such as grok-4.6. Requests go only to xAI’s fixed official API endpoint, and runs retain the model actually served and the available token usage.
The model participates through Nucovo’s conversations, routines, tasks, evidence and reviews. This gives the office a Grok-powered employee without presenting Grok Bot’s private product orchestration as an API integration.
Provider credentials stay with the trusted worker and are never put in model prompts or the isolated task runner.
Build a room around the work.
Then let the evidence speak.
Create a conversation in Teammates or prepare a routine for a project.