Caprock, in words and files you can reuse.

Caprock — the ledger of agent work

Who is waiting on you, who is stuck, how close the limit is.

A free desktop app for developers who run coding agents all day: who is waiting on you, who is stuck, how close each limit is, and what each pull request cost.

Caprock is the ledger of agent work, free, on your machine: every coding agent you run — Claude Code, Codex, OpenCode, Gemini CLI — your own terminal's sessions included; who is waiting on you, who is stuck, how close each limit is, what each pull request cost. Apache-2.0.

Caprock is a free, open-source desktop app for macOS, Windows and Linux that keeps the ledger of agent work on the developer's machine: every coding agent they run, who is waiting on them, who is stuck, how close each plan is to its limit, and what each pull request cost at list price. Claude Code, Codex, OpenCode, Gemini CLI and Hermes Agent each run in a terminal tab of their project, with worktrees, a diff to review and commit, the sessions waiting on you first. Its Pipeline and Work screens follow every branch from agent to pull request, CI, merge and release, with the agent cost of each.

Caprock is a free, open-source desktop app for macOS, Windows and Linux that keeps the ledger of agent work on the developer's machine: every coding agent they run, who is waiting on them, who is stuck, how close each plan is to its limit, and what each pull request cost. It records every session on the machine whether or not it started it, across Claude Code, Codex, OpenCode, Gemini CLI and Hermes Agent, prices each turn at list price when it is written, and joins it to the repository, pull request, CI run and release it produced. Developers run their agents side by side in it, each in a real terminal tab of its project, beside plain shells, with split panes, or keep their own terminal: the ledger is the same either way. The sidebar lists every project with its worktrees and the sessions in each, and puts the sessions waiting on you first; on macOS and Linux a permission prompt from a Claude Code session Caprock started arrives as a notification with Approve and Deny. Typing a task into the command palette opens a new agent on it, in a new git worktree, and its changes can be reviewed, committed, pushed and opened as a pull request from the app. With GitHub connected, the Work screen shows every branch on the machine at its furthest stage — agent, pull request, CI, merged, released — with the agent and model on it, the time it has taken and its agent cost at API list price. The Pipeline screen adds the next release with a progress bar per pull request, what each past release cost, and the repository's pace against its last 14 days. A red check or a merge conflict can be handed to an agent in that branch with one button. Underneath is a small daemon bound to 127.0.0.1 that records every agent session on the machine, including the ones started in a terminal, in a SQLite file on the developer's disk: cost per repository and per pull request at list price, Claude and Codex plan limits, loop alerts, and everything the agents said, searchable past Claude Code's 30-day cleanup. A phone pairs by QR code and reaches it from anywhere through an end-to-end encrypted relay that carries only ciphertext. There is no account, no cloud copy of the data and no telemetry. Caprock is Apache-2.0 and free; Premium pauses the sessions Caprock starts at a daily budget or near the plan limit, answers them from Telegram and adds a weekly report. For a team, Caprock for Teams is a self-hosted server that adds every machine up by repository, pull request and release, however the company pays for tokens, on machines it owns.

Caprock is the ledger of agent work: every coding agent on every machine, what it is doing now, what it cost per pull request however you pay, kept on machines you own. For a developer: a free, open-source desktop app that shows who is waiting on them, who is stuck and how close each limit is, with no account, hosted service or telemetry. For a team: a self-hosted server that adds every laptop up by repository, pull request and release. Apache-2.0, with an optional Premium plan. caprock.dev

Not a vendor dashboard: it prices Bedrock, Vertex, seat-plan and API sessions alike, from the transcript on the machine. Not a cockpit: it records sessions whether or not it started them, across Claude Code, Codex, OpenCode and Gemini CLI. Not a usage widget: it joins each session to its repository, pull request, CI run and release, and a team version adds machines up on a server the team runs. Not engineering-intelligence SaaS: no vendor cloud, no admin API, and the team screen counts work units, not people. Sessions that run in a vendor's cloud or inside CI are not observed; Caprock sees the machines it is installed on. Every dollar Caprock shows is an estimate at API list price, as Claude Code's own /usage is: not a bill, and on a flat plan not money saved.