Agentsview is a free, locally-installed tool that consolidates token usage and cost tracking across all AI coding agents on your machine into a single web interface.
It claims 100x faster query speed than competing tools by indexing session data locally in SQLite, and supports both individual developers and teams via Docker or cloud-backed databases.
What happened
A new tool called Agentsview launched, allowing developers to browse, search, and track token consumption and costs across multiple AI coding agents (Claude, Devin, Codex, and others) in a single local interface. The tool runs as a single binary with no accounts required and stores all data locally.
Why it matters
Developers working with multiple AI agents can now see spending across all their coding tools from one place. Agentsview claims its indexing approach delivers queries over 100x faster than tools that re-parse raw session files each run, making cost tracking practical for teams managing multiple agent instances.
What to watch
The tool is available as a macOS/Linux shell installation, Windows desktop app, Docker image, or via Homebrew. It supports remote access via SSH port-forwarding and can back data to PostgreSQL or DuckDB for team deployments, with optional semantic search across session content if connected to an OpenAI-compatible embeddings endpoint.
Agentsview is a cost-tracking tool for developers using multiple AI coding agents simultaneously. It automatically discovers sessions from Claude, Devin CLI, Codex, Forge, and OpenCode on a user's machine, syncs them into a local SQLite database, and serves a web UI at http://127.0.0.1:8080 on first run. The tool is offered as a single binary (installable via shell script on macOS and Linux, PowerShell on Windows, or Homebrew), a desktop app for macOS and Windows, or a Docker image.
Core to Agentsview's value proposition is speed: the tool claims queries are over 100x faster than competing utilities like ccusage because session data is already indexed in SQLite rather than re-parsed from raw files on each run. The agentsview usage command provides daily cost summaries (defaulting to the last 30 days), per-model breakdowns, date and agent filtering, JSON output for scripting, and timezone-aware bucketing. The tool automatically sources pricing via LiteLLM rates with an offline fallback, and calculates costs with awareness of prompt-caching economics (distinguishing cache-creation tokens from read tokens). The deprecated agentsview token-use command now also reports cost estimates.
The web interface includes a session browser with full-text search across message content, an optional semantic-search mode (when connected to an OpenAI-compatible embeddings endpoint), a token-usage and cost dashboard with per-session and per-model breakdowns, an analytics dashboard with activity heatmaps and velocity metrics, and a Recent Edits feed showing the files agents changed most recently across every session. The tool supports live updates via server-sent events as active sessions receive new messages and keyboard-first navigation. Sessions can be exported as HTML or published to GitHub Gist.
For team deployments, Agentsview can centralize sessions from multiple machines by pointing Claude and Codex session sources at S3-compatible object storage. The tool also supports Docker deployments backed by PostgreSQL or DuckDB, with examples provided for both production and development setups. Remote access via SSH port-forwarding, reverse proxies, and dev environments (exe.dev, Codespaces, Coder, WSL2) is supported by setting the --public-url flag. The tool runs with no accounts or authentication by default, though authentication can be enabled for exposed instances via --require-auth.
Agentsview addresses a growing pain point for developers working with multiple AI coding agents: fragmented cost visibility. As teams adopt Claude Code, Devin, and other AI tools in parallel, each tool typically tracks costs separately, forcing developers to manually aggregate spending across multiple dashboards or log files. The tool's core claim — that it is over 100x faster than re-parsing session files on every query — rests on its architectural choice to index session data into a local SQLite database once, then run all subsequent queries against that indexed store. This design also enables the tool to work entirely offline and without authentication, lowering barriers to adoption for developers skeptical of cloud-based cost trackers.
The depth of integration the tool attempts is notable: it aims to discover sessions from five different agent platforms, calculate costs using LiteLLM's pricing rates offline, account for prompt-caching economics (distinguishing cache-creation tokens from read tokens), and even support git-derived outcome metrics like commits and files changed. For teams, the Docker deployment examples show how to centralize sessions across multiple machines via S3 or cloud databases, suggesting the tool is designed to scale from individual developers to larger organizations.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
Silicon Motion announced a private placement of US$1 billion in aggregate principal amount of 0% convertible s…
Sundar Pichai, CEO of Alphabet and Google, announced on August 11 that the Gemini app's monthly active users (…
OpenAI's ChatGPT crossed 1 billion monthly users some time ago and hit 1 billion weekly users in July, accordi…
OpenAI's Daybreak Red (GPT-5.6 Cyber) and Daybreak Blue (GPT-5.6 Sol) are now available to eligible customers…
IBM and Together AI have signed a $240 million agreement to build and operate an Nvidia-powered AI inference c…
NVIDIA and partners released multiple open-source AI models optimized for local execution throughout August, i…
The AI news that matters, in one minute each morning.
Sign up free