Skip to content

AgentsView Documentation

AgentsView lets you browse, search, and track costs across your AI coding sessions. A background server imports recorded sessions into a SQLite database on your machine. The web interface, desktop app, command line, and APIs read that archive.

New here? The product overview explains what AgentsView is for, and the five-minute guide walks the whole loop with screenshots. The pages here explain how to use and maintain each feature.

These guides follow main and may include changes newer than the latest release, v0.46.1. Check the changelog for what each release includes.

Quick Start View on GitHub

Start here

If you want to… Read…
Install and see your sessions in under a minute Quick Start
Learn the web interface Usage Guide
See when agents ran, overlapped, and what it cost Activity
Get daily token and cost reports Token Usage & Costs
Search transcripts by meaning, not just words Semantic Search
Correct project names and worktree assignments Data
Score session health and outcomes Session Intelligence
Browse extracted, provenance-linked knowledge Recall
Give agents and scripts access to the archive MCP Server and Session API
Share sessions across machines Local Session Archive, Hosted Raw Sync, PostgreSQL Sync, ClickHouse Sync, DuckDB Mirror, Filesystem Session Sync
Sync Claude.ai or import chat exports Chat Import
Fetch Codex cloud chats Codex Cloud
Sync a remote machine over SSH SSH Remote Sync
Configure discovery, paths, and settings Configuration
Look up a command or flag CLI Reference

How the archive fits together

The daemon imports your sessions. This background server owns the writable archive and watches your agents' directories for new messages. The desktop app and commands that need the server start it when required. Some diagnostic commands can read the archive directly. See agentsview daemon.

SQLite keeps your session history. Parsed sessions live in ~/.agentsview/ with full-text indexes. Optional backends extend it: PostgreSQL or ClickHouse for a remote read-only copy and DuckDB for analytical reads. All three can receive mirrors pushed from SQLite. Alternatively, hosted raw sync can retain original provider files and parse them into PostgreSQL without a local SQLite archive or a separate push.

AgentsView architecture: agent sessions sync into SQLite with FTS5 search, served via REST API, SSE events, and embedded Svelte SPA

Use Configuration to choose source directories and how much content to retain. Use Data to manage archive size.

Privacy

Session data stays on your machine by default. The server binds to 127.0.0.1 unless you explicitly configure remote access. Data leaves the machine only for features you choose, such as hosted raw sync, a PostgreSQL or ClickHouse target, remote DuckDB access, Generated Insights, or publishing a session to GitHub. Telemetry is an anonymous, content-free daemon liveness ping plus app-open and core-action events the web UI reports through the server, and AGENTSVIEW_TELEMETRY_ENABLED=0 disables all of them.

Human and machine-readable pages

Every page has an HTML URL and a Markdown twin. For example:

  • https://agentsview.io/docs/token-usage/
  • https://agentsview.io/docs/token-usage.md

The complete machine-readable index is available at https://agentsview.io/llms.txt.