Skip to main content
Enterprise observability gives you a cross-developer view of how AI is being used across your entire organization. You see usage by developer, team, model, and project — and when something looks unusual, you can follow the trail from a metric spike all the way to the exact session that caused it. No guesswork, no waiting for a monthly bill with no context behind it.

​
Org-wide usage dashboard

The enterprise usage dashboard aggregates captured session data across every developer in your org. It is the starting point for understanding AI usage at scale.

Usage by developer

See which developers are the heaviest users and how their usage changes over time. Drill into any developer’s session list directly from the dashboard.

Usage by team

Compare activity across teams. Identify which teams are driving the most LLM traffic and which are underutilizing AI tooling.

Usage by model

Understand your model distribution. See whether teams are standardizing on approved models or using an uncontrolled mix of providers.

Usage by project

Track which projects are generating the most AI activity. Project-level attribution helps link AI investment to the work it supports.
Cross-developer analytics refresh every 15 minutes. Dashboard figures may lag up to 15 minutes behind real-time capture activity.

​
Session drill-down

The dashboard is the entry point, not the destination. Every metric and aggregation links directly to the sessions behind it.
1

Identify a signal

Spot a spike in token usage, an unusual model in the distribution, or a team whose activity has changed sharply. The dashboard surfaces these as time-series charts and ranked lists.
2

Filter to the relevant context

Apply filters by developer, team, model, date range, or project to narrow the view to the activity you want to investigate.
3

Open the session list

Click through from any chart or table row to a filtered session list scoped to that developer, team, or time range.
4

Inspect the session

Open an individual session to see the full interaction — messages, tool calls, file diffs, token counts, and model metadata. For encrypted sessions, admins with decrypt access can open the raw transcript.

​
Token usage and efficiency

Navigate to Enterprise → your org → Tokens to see a per-project token breakdown with efficiency ratios. This view helps you understand not just how many tokens you’re using, but whether that usage is working well.

​
Metrics available

A high tokens-per-line-changed ratio may indicate developers are sending unnecessary context on every request. A low cache-hit ratio may suggest prompt structure optimizations that could meaningfully reduce cost.

​
Historical backfill

When a developer joins your organization, their existing captured sessions can be backfilled into the enterprise view. This means your analytics aren’t limited to sessions captured after you set up the org — you get a full historical picture from day one.
Backfilled sessions are attributed to your org and the developer’s assigned team using the team assignment active at the time of backfill. Historical attribution reflects current team membership, not team membership at the time the session was originally captured.

​
What’s coming

The current observability release covers the org-wide usage dashboard, cross-developer analytics, session drill-down, and the Tokens tab. The following are in active development:
Break spend down by team, developer, project, and org. Set budget thresholds and get notified before costs exceed them — via email, Slack, or webhook.
Automatically flag runaway agents, token spikes, and unusual usage patterns. An anomaly review queue groups flagged activity so admins can investigate without manually watching dashboards.
Receive real-time alerts when DLP scanning flags sensitive content in captured sessions. Route alerts to Slack, PagerDuty, email, or a webhook.
Export structured event data via webhook or syslog into your existing SIEM stack for correlation with other organizational security signals.
Forward normalized cross-provider metrics and events from RCLM into your existing observability stack — dashboards, alerting, or event analysis tools — through a single export integration.
Compare Claude, OpenAI, Gemini, and other provider usage on a common reporting surface, with normalized per-token cost estimates regardless of provider-specific pricing structures.

​
How RCLM extends your existing visibility

If your developers already use Claude Code or other AI coding assistants, you may have some usage metrics available through those tools. ReclaimLLM adds the layers that standalone tool metrics cannot provide on their own.
Individual tool metrics tell you what happened in aggregate. RCLM adds the full session record so any alert or anomaly can be traced all the way back to the exact interaction that triggered it.

​
Next steps

LLM Gateway

Route gateway traffic through RCLM so it flows into the same analytics surfaces as captured sessions.

Roles & Access

Control which roles can see org-wide analytics and drill into individual sessions.