Decant CLI
Local analytics for Claude Code, Codex and five more agents: tokens, cost, time and log search in one archive
- 75 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The tool reads all local agent logs, which can contain code and pasted secrets
- Publishing the Docker port without the `127.0.0.1` prefix exposes the session archive to the network
Install
- Works in
Manual install
npx @dosu/decant@latestOne-off launch without installing. For a persistent command use `npm install --global @dosu/decant@latest` or `brew install dosu-ai/dosu/decant`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The tool reads all local agent logs, which can contain code and pasted secrets
- Publishing the Docker port without the `127.0.0.1` prefix exposes the session archive to the network
This is third-party code. Review the repository files before installing.
What it does
Decant gathers the session logs that Claude Code, Codex, Gemini CLI, OpenCode, Cursor CLI, GitHub Copilot CLI and Copilot Chat in VS Code already write on your machine, plus imported ATIF trajectories, into one SQLite archive. `npx @dosu/decant@latest` indexes the logs it finds, opens a web UI at `127.0.0.1:3000` and keeps watching for new sessions. The interface shows tokens, estimated cost, context window and agent time by agent, model, project, repository and day, along with file hotspots and tool and MCP server usage. Full-text search covers prompts, responses, tool names and their inputs. The command line offers `decant ls`, `decant search`, `decant stats --by model`, `decant economics` and session export to Markdown, JSON, trajectory-v1 and ATIF. From command history Decant builds recommendations, workflow scripts and draft instructions: SKILL.md, AGENTS.md sections and slash commands. HTML reports leave out message text but include project paths.
Who it is for. Developers and team leads who work with several coding agents at once and want to see where tokens, money and time go.
Good fit when
- The token bill grew and you need to see which projects and models spend the most
- You need to find an old session about an authorization bug by the text of the prompt
- You want to compare Claude Code and Codex on the same repositories by time and cost
- Recurring commands from history should become a skill or a script
Not a fit when
- You need cloud session logs: Decant reads only files agents write locally
- You need the exact cost from a provider invoice: prices in Decant are estimates from a model table
- You need a shared team dashboard on a server: the tool is built for local use
Example request
Show which models and projects spent the most tokens this week and find the session where we fixed authorizationLimitations
Decant reads your logs in full, and logs can hold code, paths and secrets you pasted into prompts. The archive is local, but an HTML report contains project paths and a prompt preview and should be reviewed before sharing. In Docker publish the port only as `127.0.0.1:3000:3000`: an open port exposes the archive to the network. Cost is an estimate computed from a model price table. Token accounting differs between agents, and where it is not recorded the column stays empty.
How to disable. Stop the Decant process and, if you installed the service, run `decant service uninstall`. The package is removed with `npm uninstall --global @dosu/decant`. `decant session rm <id> --yes` deletes session data and `decant db info` shows where the database lives.
FAQ
Which agents are supported?
Claude Code, Codex, Gemini CLI, OpenCode 1.2 and later, Cursor CLI, GitHub Copilot CLI and Copilot Chat in VS Code. The default log folder decides, and it can be overridden.
How do I install Decant permanently?
Three ways: `npm install --global @dosu/decant@latest`, `brew install dosu-ai/dosu/decant` or the shell installer. `decant update` upgrades it.
What goes into the HTML report?
Analytics without message text and tool bodies but with project paths. A single-session report adds the title, a prompt preview and file paths.
Can Decant run as a service?
Yes, `decant service install` starts it at login on macOS. By hand the web UI starts with `decant serve`, and `--no-open` disables opening the browser.
What is ATIF?
An agent trajectory format that Decant can import and export. `decant export 1 --as atif` saves a session in it.
Discussion
No questions or reviews yet.
Related
CLI toolsOne browser window for Claude Code, Codex, Gemini CLI and OpenCode: session status, history and requests between agents.
A local real-time dashboard for Claude Code and Codex sessions: agent activity and tool calls
A CLI that reads local Claude Code, Codex and Hermes sessions, finds repeated failures and stages fixes for manual review
A CLI and skill that gather coding-agent session history into one SQLite file the agent can search for past decisions
ClawMetry reads the session files agents already write and shows the timeline, tool calls, tokens and cost in one dashboard.
DevChain runs a team of AI coding agents on your own machine with a shared board, chat, code review and live terminals.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Decant (this card) | no | 75 | |
| ccusage | no | 19,425 | token and cost reports from logs with no full-text search or web session archive |
| Claude Code Usage Monitor | no | 8,801 | a terminal monitor of Claude Code spend in real time |
| OpenUsage | no | 246 | a terminal dashboard of quotas and limits across several tools |
README in short
The README opens with a one-command launch, then lists features, supported agents with log paths and install routes via npm, Homebrew, a script and Docker. A command reference and a description of the web UI tabs follow.