Ask an engineering director how many AI coding agents are running in their organization right now. Most can't answer. They know the team uses Cursor. Some developers switched to Claude Code last month. The infrastructure team is experimenting with Copilot Workspace. But how many sessions are active? What repositories are they touching? Which ones are blocked? Nobody knows.

Fleet closes that gap.

AI agent observability: what Fleet shows you

Fleet is a real-time AI agent observability dashboard for every coding agent operating across your engineering organization. One screen. Every agent. Every session.

When you open Fleet, you see:

  • Active sessions. How many agents are running right now, broken down by team, developer, and agent type (Cursor, Claude Code, Copilot, custom MCP agents).
  • Repository activity. Which repos agents are working in, what files they're touching, and how many changes they've proposed in the last hour, day, or week.
  • Rule violations. How many governance rules were triggered, which rules get violated most, and which agents produce the most violations.
  • Token consumption. Real-time and historical token usage attributed to teams, developers, and specific tasks. No more unexplained API bills.
  • Blocked sessions. Agents that hit a governance rule and stopped. Agents stuck in retry loops. Agents waiting on human approval.

Why multi-agent AI development needs observability

Six months ago, most teams had one or two developers experimenting with AI agents. Now multi-agent AI development is the default workflow. A 200-person engineering org might have 60-80 agent sessions running concurrently during peak hours. Each one is generating code, consuming tokens, and touching production codebases. Enterprise AI coding tools need visibility to match this scale.

Without visibility, problems stack up:

Cost surprises. A developer leaves an agent running in a retry loop over a weekend. Monday's bill is $800 for a single session. With Fleet, you'd see the blocked session in real time and get an alert after 30 minutes of inactivity.

Uncoordinated changes. Two developers on different teams both ask their agents to refactor the same shared module. Neither knows about the other. Both open PRs. The merge conflict wastes a day. Fleet shows overlapping repository activity in real time.

Governance drift. One team diligently uses governance rules. Another team doesn't. Over time, their codebases diverge in quality and consistency. Fleet's compliance view makes this visible at the team level, not just the PR level.

How Fleet works

Fleet integrates through the same Uzera SDK that powers governance rules. If your agents are already running through Uzera, Fleet is a dashboard toggle -no additional setup.

For agents not yet connected to Uzera, Fleet provides lightweight adapters:

  • Cursor: A VS Code extension that reports session metadata (active/idle, files accessed, tokens consumed) without reading code content.
  • Claude Code: A CLI wrapper that captures session lifecycle events.
  • GitHub Copilot: Integration via Copilot's enterprise API for usage analytics.
  • MCP agents: Any agent communicating over MCP gets automatic Fleet telemetry through the protocol layer.

Fleet never reads or stores your source code. It captures metadata for the AI agent audit log -which files were accessed, how many tokens were consumed, which rules were triggered -not the content of the code itself. This AI coding audit trail gives teams full accountability without compromising code privacy.

Three views for three audiences

Developer view

Your personal agent activity. How many sessions you ran today, your token spend, which governance rules you hit most often, and your code acceptance rate. It's your personal scoreboard for agent-assisted work.

Team lead view

Your team's aggregate activity. Which developers are most active with agents, which repositories are getting the most agent attention, and where your team's governance compliance stands. Useful for sprint retros and capacity planning.

Org view

The full picture. Total agent sessions, org-wide token spend trending, cross-team repository overlap, and the governance rules that get violated most across the organization. This is the view for engineering directors and VPs who need enterprise AI coding tools that show agent adoption at scale.

AI agent audit log and alerts that matter

Fleet includes configurable alerts for the situations that actually warrant attention:

  • Runaway sessions: Agent stuck in a loop for more than N minutes. Configurable per team.
  • Spend threshold: Daily or weekly token spend exceeds a budget. Per-developer or per-team.
  • Governance spike: Sudden increase in rule violations, which often signals a new developer or a new agent that hasn't been configured with your rules.
  • Overlap detection: Multiple agents actively modifying the same files across different branches.

Ship date: today

Fleet is included in Uzera Team and Enterprise plans starting today. If you're on the Starter plan, you get the developer view -your personal agent activity and token spend.

For teams already using Uzera, Fleet is live in your dashboard. For new teams, you can set up Fleet alongside your first governance rules during onboarding.

The age of unmonitored agent activity is over. You wouldn't run 80 cloud services without a monitoring dashboard. You shouldn't run 80 coding agents without one either.