agentsview-local-agent-analytics

name	agentsview-local-agent-analytics
description	Local-first session search, analytics, and cost tracking for AI coding agents with SQLite storage
triggers	["track my AI agent token usage and costs","search through my coding agent sessions","analyze my AI coding assistant activity","view analytics for my agent conversations","export my Claude Code or Cursor sessions","set up local agent session tracking","monitor my AI coding agent spending","get daily cost breakdown for coding agents"]

agentsview Local Agent Analytics

Skill by ara.so — AI Agent Skills collection.

agentsview is a local-first tool for browsing, searching, and tracking costs across 20+ AI coding agents including Claude Code, Codex, Cursor, Copilot, and more. Everything runs locally with no accounts or cloud dependencies. Built in Go with SQLite storage, it provides a web UI for session browsing, full-text search, token usage analytics, and cost tracking.

Installation

Quick Install

# macOS / Linux
curl -fsSL https://agentsview.io/install.sh | bash

# Windows
powershell -ExecutionPolicy ByPass -c "irm https://agentsview.io/install.ps1 | iex"

Homebrew (macOS)

brew install --cask agentsview

Docker

docker run --rm -p 127.0.0.1:8080:8080 \
  -v agentsview-data:/data \
  -v "$HOME/.claude/projects:/agents/claude:ro" \
  -v "$HOME/.codex/sessions:/agents/codex:ro" \
  -e CLAUDE_PROJECTS_DIR=/agents/claude \
  -e CODEX_DIR=/agents/codex \
  ghcr.io/kenn-io/agentsview:latest

From Source

git clone https://github.com/kenn-io/agentsview.git
cd agentsview
go build -o agentsview ./cmd/agentsview
./agentsview serve

Core Commands

Server Management

# Start foreground server (blocks terminal)
agentsview serve

# Start background server
agentsview serve --background

# Check server status
agentsview serve status

# Stop background server
agentsview serve stop

# Serve on specific port
agentsview serve --port 9090

# Enable authentication (required for non-loopback)
agentsview serve --require-auth

Token Usage & Cost Tracking

# Daily cost summary (last 30 days)
agentsview usage daily

# Per-model breakdown
agentsview usage daily --breakdown

# Filter by agent and date range
agentsview usage daily --agent claude --since 2026-04-01 --until 2026-04-30

# All-time summary
agentsview usage daily --all

# JSON output for scripting
agentsview usage daily --all --json

# One-line statusline for shell prompts
agentsview usage statusline

# Per-session usage
agentsview session usage <session-id>
agentsview session usage <session-id> --format json

Session Analytics

# Human-readable stats (last 28 days)
agentsview stats

# JSON output with custom date range
agentsview stats --format json --since 2026-04-01 --until 2026-04-15

# Filter by agent
agentsview stats --agent claude

# Include git metrics (slow, opt-in)
agentsview stats --include-git-outcomes

# Include GitHub PR metrics (requires gh CLI)
agentsview stats --include-github-outcomes

Database Operations

# Sync agent sessions into SQLite
agentsview sync

# Export to DuckDB
agentsview duckdb push --full

# Serve DuckDB mirror read-only
agentsview duckdb serve

# PostgreSQL mode
agentsview pg serve

Configuration

Environment Variables

# Agent session directories (auto-detected by default)
export CLAUDE_PROJECTS_DIR="$HOME/.claude/projects"
export CODEX_DIR="$HOME/.codex/sessions"
export CURSOR_DIR="$HOME/.cursor/sessions"
export FORGE_DIR="$HOME/.forge"
export COPILOT_DIR="$HOME/.copilot"

# Database
export AGENTSVIEW_DATA_DIR="$HOME/.agentsview"
export AGENTSVIEW_PG_URL="postgres://user:pass@host:5432/agentsview?sslmode=require"
export AGENTSVIEW_DUCKDB_URL="quack:https://duckdb.example.com"
export AGENTSVIEW_DUCKDB_TOKEN="your-quack-token"

# Server
export AGENTSVIEW_PORT="8080"
export AGENTSVIEW_PUBLIC_URL="http://127.0.0.1:8080"

Remote Access Setup

For SSH port-forwarding, reverse proxies, or remote dev environments:

# Browser opens http://127.0.0.1:18080 via SSH tunnel
agentsview serve --public-url http://127.0.0.1:18080

# Codespaces, exe.dev, etc.
agentsview serve --public-url https://your-workspace.exe.dev

# Trust additional origins
agentsview serve --public-origin https://dev.example.com,https://staging.example.com

Docker Compose Production Setup

# docker-compose.prod.yaml
version: '3.8'

services:
  agentsview:
    image: ghcr.io/kenn-io/agentsview:latest
    ports:
      - "127.0.0.1:8080:8080"
    volumes:
      - agentsview-data:/data
      - ~/.claude/projects:/agents/claude:ro
      - ~/.codex/sessions:/agents/codex:ro
      - ~/.cursor/sessions:/agents/cursor:ro
    environment:
      - CLAUDE_PROJECTS_DIR=/agents/claude
      - CODEX_DIR=/agents/codex
      - CURSOR_DIR=/agents/cursor
    restart: unless-stopped

volumes:
  agentsview-data:

Run with:

docker compose -f docker-compose.prod.yaml up -d

REST API

Get Session Usage

# Fetch per-session token and cost data
curl http://127.0.0.1:8080/api/v1/sessions/<session-id>/usage

# Example response
{
  "session_id": "abc123",
  "agent": "claude",
  "project": "my-app",
  "total_output_tokens": 15420,
  "peak_context_tokens": 98765,
  "has_token_data": true,
  "cost_usd": 0.42,
  "has_cost": true,
  "models": ["claude-3.5-sonnet"],
  "unpriced_models": [],
  "server_running": true
}

Settings & Configuration

# Get settings
curl http://127.0.0.1:8080/api/v1/settings

# Search sessions (FTS5)
curl "http://127.0.0.1:8080/api/v1/search?q=refactor+authentication"

Common Patterns

Daily Cost Monitoring in Shell Prompt

Add to .bashrc or .zshrc:

# Show today's AI agent cost in prompt
function agent_cost() {
  agentsview usage statusline 2>/dev/null || echo "agentsview offline"
}

PS1='$(agent_cost) $ '

Automated Daily Reports

#!/bin/bash
# daily-agent-report.sh

DATE=$(date +%Y-%m-%d)
REPORT_FILE="$HOME/reports/agent-usage-$DATE.json"

agentsview usage daily --all --json > "$REPORT_FILE"
agentsview stats --format json >> "$REPORT_FILE"

echo "Report saved to $REPORT_FILE"

CI/CD Cost Tracking

# .github/workflows/agent-costs.yml
name: Track Agent Costs

on:
  schedule:
    - cron: '0 0 * * *' # Daily at midnight

jobs:
  track:
    runs-on: ubuntu-latest
    steps:
      - name: Install agentsview
        run: curl -fsSL https://agentsview.io/install.sh | bash
      
      - name: Mount agent sessions
        run: |
          mkdir -p ~/.claude/projects
          # Sync from cloud storage or artifact
      
      - name: Generate report
        run: agentsview usage daily --all --breakdown --json > costs.json
      
      - name: Upload artifact
        uses: actions/upload-artifact@v3
        with:
          name: agent-costs
           

Querying the SQLite Database Directly

# Open the database
sqlite3 ~/.agentsview/sessions.db

# Query recent sessions
SELECT 
  id, 
  agent, 
  project_name,
  started_at,
  message_count
FROM sessions
WHERE started_at > datetime('now', '-7 days')
ORDER BY started_at DESC
LIMIT 10;

# Get token stats by model
SELECT 
  model,
  SUM(total_tokens) as total,
  AVG(total_tokens) as avg_per_session
FROM session_usage
GROUP BY model
ORDER BY total DESC;

Export Sessions Programmatically

package main

import (
    "encoding/json"
    "fmt"
    "net/http"
)

type SessionUsage struct {
    SessionID          string   `json:"session_id"`
    Agent              string   `json:"agent"`
    TotalOutputTokens  int      `json:"total_output_tokens"`
    CostUSD            float64  `json:"cost_usd"`
    HasCost            bool     `json:"has_cost"`
}

func fetchSessionUsage(sessionID string) (*SessionUsage, error) {
    url := fmt.Sprintf("http://127.0.0.1:8080/api/v1/sessions/%s/usage", sessionID)
    resp, err := http.Get(url)
    if err != nil {
        return nil, err
    }
    defer resp.Body.Close()

    var usage SessionUsage
    if err := json.NewDecoder(resp.Body).Decode(&usage); err != nil {
        return nil, err
    }

    return &usage, nil
}

func main() {
    usage, err := fetchSessionUsage("your-session-id")
    if err != nil {
        panic(err)
    }

    fmt.Printf("Agent: %s\n", usage.Agent)
    fmt.Printf("Tokens: %d\n", usage.TotalOutputTokens)
    if usage.HasCost {
        fmt.Printf(, usage.CostUSD)
    }
}

Background Server with Systemd

# ~/.config/systemd/user/agentsview.service
[Unit]
Description=agentsview local agent analytics
After=network.target

[Service]
Type=simple
ExecStart=/usr/local/bin/agentsview serve --port 8080
Restart=on-failure
RestartSec=5s
Environment="AGENTSVIEW_DATA_DIR=%h/.agentsview"

[Install]
WantedBy=default.target

Enable and start:

systemctl --user enable agentsview.service
systemctl --user start agentsview.service
systemctl --user status agentsview.service

Supported Agents & Session Paths

Agent	Default Session Directory
Claude Code	~/.claude/projects/
Codex	~/.codex/sessions/
Cursor	~/.cursor/sessions/
Copilot CLI	~/.copilot/
Forge	~/.forge/
Amp	~/.local/share/amp/threads/
Claude Cowork	~/Library/Application Support/Claude/local-agent-mode-sessions/ (macOS)
Antigravity	~/.gemini/antigravity/
Cortex Code	~/.snowflake/cortex/conversations/

All paths are auto-detected. Override with environment variables if needed.

Troubleshooting

Server Returns 403 Forbidden

Symptom: API requests return 403 Forbidden when accessing through SSH tunnel, reverse proxy, or remote dev environment.

Cause: Host header validation fails when the browser's Host header doesn't match the server's expected origin.

Solution:

# Set public URL to match browser origin
agentsview serve --public-url http://127.0.0.1:18080

# Or trust multiple origins
agentsview serve --public-origin https://dev.example.com,https://codespaces.github.dev

Sessions Not Appearing

Symptom: Agent sessions don't show up in the UI.

Cause: Session directory not mounted (Docker) or not auto-detected.

Solution:

# Verify session directory exists
ls ~/.claude/projects/

# Docker: ensure volume is mounted
docker run -v "$HOME/.claude/projects:/agents/claude:ro" \
  -e CLAUDE_PROJECTS_DIR=/agents/claude \
  ghcr.io/kenn-io/agentsview:latest

# Manual sync
agentsview sync

Cost Data Missing

Symptom: has_cost: false or cost_usd: 0.

Cause: Model pricing not available in LiteLLM database or no token data recorded.

Solution:

# Check if token data exists
agentsview session usage <id> --format json | jq '.has_token_data'

# Update agentsview to latest version
curl -fsSL https://agentsview.io/install.sh | bash

Background Server Not Starting

Symptom: agentsview serve --background fails or exits immediately.

Solution:

# Check logs
cat ~/.agentsview/serve.log

# Verify port not in use
lsof -i :8080

# Try foreground to see errors
agentsview serve

Slow Git Outcome Metrics

Symptom: agentsview stats --include-git-outcomes hangs or takes too long.

Cause: Large repositories or missing git history.

Solution:

# Omit git metrics (default)
agentsview stats

# Or limit date range
agentsview stats --include-git-outcomes --since 2026-06-01

Docker Permission Issues

Symptom: Root-owned files in ~/.agentsview after Docker run.

Solution:

# Use named volume instead of bind mount
docker run -v agentsview-data:/data ghcr.io/kenn-io/agentsview:latest

# Or pre-create with correct ownership
mkdir -p ~/.agentsview
docker run -v "$HOME/.agentsview:/data" ghcr.io/kenn-io/agentsview:latest

Advanced Usage

PostgreSQL Backend

# Set PostgreSQL URL
export AGENTSVIEW_PG_URL="postgres://user:pass@host:5432/agentsview?sslmode=require"

# Start PostgreSQL mode
agentsview pg serve

# Docker
docker run -e PG_SERVE=1 \
  -e AGENTSVIEW_PG_URL="$AGENTSVIEW_PG_URL" \
  ghcr.io/kenn-io/agentsview:latest

DuckDB Mirror

# Export SQLite to DuckDB
agentsview duckdb push --full

# Serve DuckDB read-only
agentsview duckdb serve

# Quack remote access
QUACK_TOKEN="$(openssl rand -base64 32)"
agentsview duckdb quack serve \
  --bind quack:0.0.0.0:9494 \
  --token "$QUACK_TOKEN" \
  --allow-insecure

# Connect to remote Quack
export AGENTSVIEW_DUCKDB_URL="quack:https://duckdb.example.com"
export AGENTSVIEW_DUCKDB_TOKEN="$QUACK_TOKEN"
agentsview duckdb serve

Custom Timezone for Reports

# Use specific timezone for daily bucketing
agentsview usage daily --timezone America/New_York

# Or set environment variable
export TZ=Europe/London
agentsview usage daily

Keyboard Shortcuts (Web UI)

• j / k — Navigate sessions
• [ / ] — Previous/next session
• Cmd+K / Ctrl+K — Open search
• ? — Show all shortcuts
• Esc — Close dialogs/modals

License

MIT