Command line
Every kelve command, as kelve --help gives it. Your agent uses its tools instead. These commands are for you, and for agents without MCP. kelve <command> --help says more about each. kelve --version says which version you have.
Show something
| Command | What it does |
|---|---|
kelve open [path] | Opens a folder or a file in kelve, in your browser, starting kelve if it isn’t running. . is this folder. With no path, it opens kelve itself. |
kelve diff | Opens what you haven’t committed yet. |
kelve diff <branch> | Opens this branch, against where it left that one. For example, kelve diff main. |
kelve diff <commit> | Opens one commit. For example, kelve diff 3f2a9c1. |
kelve worktree | Shows what an agent working in a worktree has done so far. Run it in the worktree. |
Each opens with a new link that signs your browser in, once.
kelve worktree takes three flags:
--baseis the branch the worktree left (mainunless you say).--saidis what’s been done, in a line.--agentis the agent as it names itself.
Run kelve
| Command | What it does |
|---|---|
kelve serve | Runs kelve in this terminal, showing what it does, and opens it in your browser. kelve open starts it in the background instead. |
kelve status | Says whether kelve is running, where, and which version, and when a newer one is out. |
kelve stop | Stops kelve, once the work it’s doing is done. |
kelve update | Updates kelve to its newest version, checked against the release’s checksums, and restarts it if it’s running. Update and uninstall says more. |
Set it up
| Command | What it does |
|---|---|
kelve setup | Sets kelve up, as the installer does. It connects your agent, adds its hook for your notes and lessons, and sets how much kelve marks. In a terminal it asks. Run by your agent, it leaves your agent to choose how much kelve marks. |
kelve setup --agent <agent> | Connects that agent: claude, codex, cursor or gemini. Without it, kelve connects the agent running it, or the one installed. |
kelve setup --level <level> | Sets how much kelve marks, as kelve marks set does. |
kelve marks set <level> | Sets how much kelve marks: new (every idea), learning (all but the basics), working (advanced and specialist ideas) or deep (specialist ideas only). The rest of what kelve marks stays as it was. |
kelve hook install <agent> | Adds kelve’s one hook to your agent’s settings, leaving the rest as it was: claude, codex, cursor, copilot, gemini, antigravity, kiro, droid, qwen, qoder or augment. |
kelve hook remove <agent> | Takes kelve’s hook out, and only that. |
Leave notes for your agent has what the hook does.
For your agent
| Command | What it does |
|---|---|
kelve mcp | kelve’s tools for your agent, over MCP. Your agent starts it itself. |
kelve job | Takes kelve’s next job, as JSON, for an agent without MCP. |
kelve job --for <number> | Takes the next job for one thing open in kelve, by its number. |
kelve job done <job> < answer.json | Hands in a job’s answer, as JSON on standard input. |
kelve kept | What’s kept for your agent to do, the most wanted first. |
kelve notes | Every note waiting for your agent, in the project it’s in. |
kelve notes --waiting | One line if notes are waiting, and nothing otherwise. |
kelve notes done <id> <version> "<did>" | Says a note is done, and what your agent did, in a line. The id and version are as kelve notes gave them. |
kelve moment --agent <agent> --on <event> | What kelve’s hook runs: reads your agent’s event, and prints the notes waiting, the lesson that helps now, or nothing. |
kelve job and kelve job done take three flags:
--agentis the agent as it names itself.--modelis the model it runs on.--foris the number of one thing open in kelve.