Agent Miner

If you're looking to integrate an AI tool into Agent Miner navigate to Docs

Agent Miner is Bloomfilter’s workspace for understanding how AI agent sessions are being used across a portfolio. It gives teams a clear view of who is using agent tools, which sources are active, how much activity is occurring, what that usage costs, and whether sessions are linked to real work items.

For new users and clients, the value of Agent Miner is simple: it turns agent activity into something visible, measurable, and easier to manage. Instead of treating AI usage as isolated conversations, Agent Miner organizes sessions into operational records that can be reviewed by portfolio, user, source, task, activity and time period.

Business Value:

Agentic Mining gives engineering leaders and finance teams full visibility into how AI agents are being used across workflows; tracking session activity, task-level impact, and cost transparency in real time, enabling organizations to maximize the ROI of their AI investments while keeping usage accountable and spend optimized.

Why It’s Valuable:

• Full Agent Visibility: See every coding-agent session across your tooling stack in one place,who ran it, how long it ran, and what it touched.
• Token & Cost Accountability: Track input, output, and cached tokens alongside API cost per session to keep AI spend transparent and defensible.
• Behavior Auditing: Replay turn-by-turn agent activity, prompts, thinking, tool calls, file reads, and edits; to validate quality and catch risky actions.
• Adoption Insights: Identify power users, idle licenses and emerging usage patterns to guide rollout, training, and license allocation.
• Linked to Delivery Work: Tie agent sessions back to the tasks and initiatives they support so AI investment maps cleanly to business outcomes.

Sessions

A live ledger of every agent coding session captured by Bloomfilter. Each row represents one session from a connected source and surfaces the user, work volume, token economics and recency of activity.

Why It Helps:

• Spot the heaviest agent users and most expensive sessions at a glance.
• Audit unusual sessions; very high turn counts, large token reads, or long-running activity before they become a cost or quality issue.
• Confirm that licensed seats are actually being used, and reclaim seats that aren’t.

Use Case:

Engineering managers reviewing weekly AI utilization, FinOps leads validating API spend, and security/compliance reviewers sampling sessions for audit evidence.

Agent Sessions

A detailed session log capturing all AI agent interactions within the platform.

Tracked Metrics:

• Session ID: Unique identifier for each agent interaction session
• Session Start: Timestamp of when the agent session began
• Agent Source: AI provider (Claude, Cursor, or other AI tools)
• Events: Total number of events within the session
• User Prompts: Count of user-initiated prompts
• Tool Calls: Number of tool/function calls made by the agent
• User: Team member who initiated the session

Key Capabilities:

• Comprehensive session history with sortable columns
• Filter sessions by individual users or view all users
• Track sessions across different AI agent sources
• Analyze engagement patterns through event and prompt counts
• Monitor tool usage frequency per session
• Date range filtering for historical analysis
• Export session data for external reporting

Session Detail

A full forensic view of a single agent session. The detail page is split into three zones: the Overview KPI cards (top), the Session Trace timeline (middle), and the per-Turn breakdown (bottom).

Why It Helps:

• Replay exactly what the agent did, in order, with the prompts, thinking, and tool calls visible.
• Validate that agent behavior matches policy—no unexpected file writes, no risky shell commands, no data exfiltration patterns.
• Diagnose runaway cost: pinpoint the turn or tool call that consumed the most tokens.

Use Case:

Security and engineering reviewers auditing a flagged session, leads coaching a developer on agent usage, or FinOps investigating a high-cost session.

Overview KPIs

• User: The developer who initiated the session.
• Started At: Absolute UTC timestamp of session start. Use this to correlate with deploys, incidents or calendar context.
• Tokens: Total tokens consumed by the session, with a secondary line for cache tokens. Cache tokens indicate context reuse and lower effective cost.
• API Cost: Estimated dollar cost for the session.

Status Indicator: Next to the session ID at the top, a colored dot shows whether the session is ACTIVE (still running / recently active) or completed.

Financials Tab

The Financials view provides cost analysis and budget management capabilities for AI agent usage, showing token consumption and associated costs.

Cost and Token Analysis

Multiple visualization options for analyzing AI agent costs and token consumption across different dimensions.

Analysis Views:

• Tasks: Cost breakdown by individual tasks
• Users: Token usage per user
• Initiatives: Cost tracking by initiative
• Epics: Cost aggregation by epic-level work

Visualization Options:

• Line Chart: Trend analysis over time
• Bar Chart: Period-by-period comparison

Key Capabilities:

• Toggle between Cost and Tokens metrics
• Filter by agent source (Cursor, Claude, or other agents)
• Time-series analysis showing cost trends over selected periods
• Multi-select filtering for tasks, users, epics, or initiatives
• Date range selection for custom reporting periods
• Export financial data for budget planning and reporting
• Identify high-cost areas for optimization opportunities

Dashboard Tab

Visualizes AI adoption and behavior trends by tracking active users over time, sessions by source, average turns per session and error rates to monitor program health, support license renewal decisions and identify enablement opportunities.

Key Consideration: Both tabs respect the same global filters (date range, team, user) as the Sessions list, so you can scope an executive view and a detailed view to the same window.