Fleet Monitoring

Monitor connected agents, detect contract drift, and track governance coverage across your fleet.

Fleet Page

Dashboard > Agents.

Fleet monitoring showing connected agents

Summary Cards

Four cards at the top:

Card	What it shows
Total Agents	Count of registered agents
Coverage	Percentage of tools covered by contracts across all agents
Enforced	Number of agents running in enforce mode
Ungoverned	Number of agents with at least one ungoverned tool

Agent Table

Below the summary cards, a sortable/filterable table of all agents:

Agent ID
Environment
Bundle name
Contract status (current / drift / unknown)
Last seen timestamp
Coverage percentage

Click any agent to open its detail page.

The right sidebar lists ungoverned tools across the fleet, sorted by agent count. Click a tool name to filter the agent table to agents using that tool without a contract.

The Events page provides a three-panel Datadog-style layout: faceted filter sidebar, event list with histogram, and detail panel. Filter by agent, tool, verdict, mode, or contract name. Time window selector ranges from 15 minutes to 30 days.

Drift Detection

Drift detection compares each agent's reported contract version against the currently deployed bundle.

How It Works

When an agent connects via SSE, it reports its policy_version (the SHA-256 revision hash of its loaded bundle)
The console compares this against the revision_hash of the latest deployed bundle for that environment
Result:

Status	Meaning
Current	Agent's contract version matches the deployed bundle
Drift	Agent's contract version doesn't match -- it's running an outdated bundle
Unknown	No bundle has been deployed for this environment yet

Resolving Drift

Drifted agents typically resolve automatically when they receive the next SSE push. If an agent stays in drift:

Check the agent's SSE connection (is it connected?)
Check if the agent reconnected after the deployment
Manually trigger a redeployment from the Contracts page

Coverage Analysis

Coverage tells you which tools are governed by contracts and which are not.

Per-Agent Coverage

Each agent's tools are classified into three categories:

Category	Meaning
Enforced	A contract in enforce mode covers this tool
Observed	A contract in observe mode covers this tool (logged but not enforced)
Ungoverned	No contract covers this tool at all

Coverage percentage = (enforced + observed) / total tools.

Coverage Sources

Coverage data comes from one of two sources, shown as a badge on each tool:

Source	When	Badge
Console	Agent has a console-deployed bundle (via compositions or direct upload)	Default -- no extra badge
Local	Agent runs Gate with local contracts and sends an `agent_manifest` alongside events	Blue "local" badge

Gate agents that govern tools locally (via gate init with a YAML file) push their contract manifest to the console during event ingestion. The console stores this manifest and uses it as a fallback coverage source when no console-deployed bundle exists for the agent. This means fleet coverage is accurate even for agents that don't use console-managed contracts.

Fleet Coverage

GET /api/v1/agents/fleet-coverage

Aggregated coverage across all agents. Returns total agents, fleet-wide coverage percentage, and the list of ungoverned tools with agent counts.

Each agent entry in the coverage response includes two additional fields:

Field	Type	Description
`last_seen`	ISO datetime (nullable)	Timestamp of the most recent tool call event from this agent. `null` if no events have been recorded.
`event_count_24h`	integer	Total tool call events from this agent in the last 24 hours.

fully_enforced semantics. An agent is counted as fully enforced only when fully_enforced > 0 — it must have at least one tool under enforcement. An agent with zero tools under any contract (where with_ungoverned === 0 only because it has no tools at all) is not counted as fully enforced.

Fleet coverage is cached in Redis for 60 seconds.

Agent Detail Page

Dashboard > Agents > click an agent.

Agent detail page with coverage and analytics

URL: /dashboard/agents/:agentId

Coverage Tab

Tool-by-tool breakdown showing:

Tool name
Coverage status (enforced / observed / ungoverned)
Contract name (if covered)
Mode (enforce / observe)

Analytics Tab

Agent analytics showing tool call patterns

Time-series data for the agent:

Tool call volume
Deny rate
Approval request count

History Tab

Agent contract change history

Timeline of contract changes and drift events:

When bundles were assigned
When drift was detected
When drift was resolved

Time Windows

All time-based views support these windows:

Window	Description
1h	Last hour
6h	Last 6 hours
24h	Last 24 hours
7d	Last 7 days
30d	Last 30 days

API Endpoints

Endpoint	Auth	Description
`GET /api/v1/agents/status`	Cookie	Live status of all connected agents (with drift check)
`GET /api/v1/agents/{agent_id}/coverage`	Cookie	Per-agent tool coverage
`GET /api/v1/agents/{agent_id}/history`	Cookie	Agent contract change timeline
`GET /api/v1/agents/fleet-coverage`	Cookie	Fleet-wide coverage summary

Next Steps

Connecting Agents -- connect more agents to the fleet
Agent Assignment -- route bundles to agents with rules