AgentsView is a local-first desktop and web application designed for browsing, searching, and analyzing AI coding agent sessions. Created by Kenn Software LLC, it enables developers to visualize what their agents have done across multiple projects, providing a centralized interface to gain insights into agent-assisted development workflows. It focuses on privacy and data ownership by keeping all session data stored locally on your machine, though it offers an optional PostgreSQL sync feature for teams and multi-machine setups.

Functionality includes indexing session logs into a searchable SQLite database, which allows for full-text search across entire conversation histories, including user prompts, assistant responses, thinking blocks, and tool calls. The application also provides an analytics dashboard, performance metrics, and cost-reporting tools, helping users understand their usage patterns and agent efficiency.

Some of the key features are:

• Cross-Agent Support: Works with over 29 AI coding agents, including Claude Code, Codex, Copilot, Cursor, Gemini, and OpenHands.
• Full-Text Search: Provides FTS5-powered searching across all message content, tool calls, and assistant thinking blocks.
• Token & Cost Analytics: Offers a sub-second CLI and web-based dashboard for daily spend reports across multiple agents.
• Local-First Architecture: Runs on a local SQLite database, ensuring no cloud account or external service dependency is required.
• Session Intelligence: Generates health scores, outcome classifications, and signal panels for quick assessment of session quality.
• Live Syncing: Watches local session directories and streams new activity to the interface in real time.
• Chat Import: Supports importing historical conversation data from Claude.ai and ChatGPT into the same local archive.
• Extensible Insights: Provides AI-generated daily summaries and multi-day analyses of coding patterns.

AgentsView works by monitoring configured session directories for changes, parsing JSONL files associated with various agent formats, and building a structured SQLite database. Users can run the application as a standalone desktop app or as a web server via the CLI, with support for browser-based navigation, command palettes, and keyboard-driven workflows for efficient session browsing. The CLI also provides a range of commands for programmatic data access, reporting, and maintenance.

Some common use cases include:

• Debugging Agent Performance: Analyzing why an agent failed a specific task by reviewing the full transcript, thinking blocks, and tool outputs.
• Budget Management: Tracking daily and monthly spending on various AI agent models to stay within project budgets.
• Knowledge Retrieval: Searching through months of past conversations to find a specific code refactoring design decision or function implementation.
• Workflow Optimization: Reviewing velocity metrics and tool usage heatmaps to improve personal coding productivity with AI.
• Team Collaboration: Syncing session data to a shared PostgreSQL instance so team members can review shared coding progress and agent behavior.