What the plugins add
LokalBot keeps a library on your computer: meetings and their action items, the people in them, the folders you chose under Files, and what you shared about yourself. The plugins teach your agent when and how to use it. Both have the same four skills:
- Brief. A short catch-up on your own work, with sources: what's in flight, what you owe and to whom, recent decisions, and which LokalBot sources are off.
- Recall. Finds what you said, saw, agreed, or saved, picks the source that fits, and cites it.
- Setup. Installs LokalBot and sets it up with you. You allow access, approve changes to what LokalBot captures, and grant system permissions in LokalBot; the agent does the rest from the terminal. The full journey is on a Mac; on Windows and Linux, the agent walks you through the install and then uses the same commands.
- The
lokalbot-cliskill. Tells the agent to check LokalBot at the start of a session on your own work and before asking you about past work, decisions, or commitments, and explains every command.
In Claude Code they're commands: /lokalbot:brief, /lokalbot:recall what did we decide about pricing?, and /lokalbot:setup. In Codex they're the skills lokalbot-brief, lokalbot-recall, and lokalbot-setup, which Codex uses when a request fits, such as “catch me up on what I owe people”.
Install in Claude Code
# Inside Claude Code
/plugin marketplace add stevyhacker/lokalbot
/plugin install lokalbot@lokalbot
/reload-plugins
The first command adds LokalBot's repository as a plugin marketplace, and the second installs the plugin.
Install in Codex
$ codex plugin marketplace add stevyhacker/lokalbot --sparse .agents/plugins --sparse Distribution/openai-plugin $ codex plugin add lokalbot@lokalbot
The --sparse options fetch only the plugin, not the whole repository. Then ask Codex to set up LokalBot, or to catch you up.
Before the agent can read anything
- Install LokalBot. The plugins work through
lokalbot-cli, which comes with the app. The setup skill can install it with you, or you can download it. The Windows and Linux apps are in alpha; see the system requirements. - Turn on access. On a Mac, in Settings → External Agents, turn on Allow external agents to read your LokalBot data, then tick Meetings, What I shared about myself, or Files. On Windows and Linux, each has its own switch under External agents in Settings → Privacy & Data. Everything is off until you turn it on.
- Put the CLI on your PATH. The plugins run
lokalbot-cliby name. Install for your coding agent…, in Settings → External Agents → Agent CLI on a Mac or Settings → Advanced → Agent CLI on Windows and Linux, puts it in~/.local/binon a Mac or Linux. On Windows it goes in%LOCALAPPDATA%\Programs\LokalBot\bin; add that folder to your PATH. The setup skill asks before doing this for you.
Screen memory needs the MCP server
Neither plugin starts LokalBot's MCP server, so screen memory, which only that server serves, isn't part of them. To add it, connect the server yourself and tick Screen memory in LokalBot:
# Claude Code $ claude mcp add lokalbot -- lokalbot-cli mcp # Codex $ codex mcp add lokalbot -- lokalbot-cli mcp
How to let Claude Code or Codex search your meeting notes lists every MCP tool and what each permission opens.
What the agent can and can't do
- It reads only what you allowed. When a source is off, the agent tells you where to turn it on instead of working around it.
- It never changes your meetings or your files. Those changes happen in the LokalBot app.
- Settings need their own switch. With Allow external agents to change LokalBot settings on, the agent can apply your preferences. Anything that widens what LokalBot captures, reads, records, or sends, and every API key, waits for your approval in LokalBot.
- Your agent's provider sees what it reads. LokalBot doesn't upload your library, but Claude Code and Codex send what a command returns to Anthropic or OpenAI under their own terms. The skills retrieve only what a question needs; allow only what you'd paste in yourself.