Agentlytics

Why it's ranked here

Its useful distinction is consolidation without sending the personal dashboard data elsewhere. Searchable conversations, project breakdowns, cost estimates, and editor comparisons address practical questions. The relay adds genuine team utility, although its trusted-network design and uneven source coverage narrow the audience.

What's good

The local workflow needs one package-runner command and keeps its cache on the machine. Session detail includes messages, models, token counts, and tool calls. Costs can be grouped by editor, model, project, and month. Relay participants choose which projects to share.

Tradeoffs

The documented runtime supports macOS only. Devin, Devin Next, and Antigravity must be running during scans. Several adapters lack complete token, model, or tool data. Cost figures are estimates tied to recognized model names. Relay is intended for trusted local networks, with password protection optional.

How to use it well

It suits developers who move among several coding agents and want one searchable record for personal analysis. Run collection periodically, compare editors by project, and inspect expensive sessions before changing habits. Use relay for selected team context on a trusted network. It does not provide LLM-generated insights or PDF and CSV reporting.

Observed

Primary language

JavaScript, using CommonJS modules

Packaging

npm package with an agentlytics command; runnable through npx, pnpm dlx, yarn dlx, or bunx

Runtime

Node.js 20.19 or newer, with additional explicit support for Node.js 22.12 or newer

Documented platform

macOS; Linux and Windows support appears on the roadmap

Interfaces

CLI, local browser dashboard, REST API, relay REST API, and relay MCP server

Storage

Separate local SQLite databases for personal cache data and relay data

License

Conflicting metadata: README.md states MIT, while package.json declares ISC