Imogen Labs
Start a project

Mission Control

See what the operators are doing.

Internal

Running agents across sessions creates an operations problem of its own. We use Mission Control as an internal console for following that activity. Our work is the operational setup and integration around the dashboard, including the memory files and workflows it brings into view.

Inside Mission Control
  1. Agent activitySession records and scheduled jobs expose ongoing work.
  2. WorkspaceMemory references and daily notes preserve operational context.
  3. ConsoleThe internal dashboard brings activity, usage and health into view.
  4. SupervisionPeople inspect the evidence within explicit operational permissions.

The problem

An agent can be working while its operator has little visibility into what it is doing or what context it can see. Session histories, daily notes, scheduled jobs and system health sit on different surfaces. Following only the chat misses the operational picture; following only the machine misses the work.

Who it’s for

We use it for people supervising ongoing agent work who need to see sessions, usage, cost and machine health together. It is internal infrastructure in our company inventory, rather than a separate commercial dashboard we ask customers to buy.

What we built

We integrated an existing open-source agent dashboard into our operations. The underlying dashboard supplies session views, usage and cost reporting, memory-file browsing, scheduled-job visibility and system health. We connect that surface to our operator workspace and the records our memory method maintains. Our contribution is the internal operational setup and integration.

How it works

The dashboard reads the configured workspace and agent session records. Its memory view exposes durable context and daily notes alongside operational activity. The result is a place to inspect whether the work and its recorded context agree, with authentication around the administrative interface.

What was hard

Visibility and authority have different boundaries. A console that can show a service may also expose controls that change it, so we pair operational access with the same explicit permissions and human decisions used elsewhere in our delivery system. The dashboard helps supervision; it is not proof that an agent's report is correct or permission to run unattended destructive work.

Have a similar problem?

Let’s build
what comes next.

Start a project