Free while in beta
Install mla
mla runs inside your coding agent, not your terminal. Paste this prompt into Claude Code or Codex and it installs mla and sets up this repo for you. It moves fast and it is free while in beta.
Please install and set up mla (the Meetless CLI) for this repository. Read the instructions at https://meetless.ai/install and follow the steps for whichever agent you are, Claude Code or Codex. Some steps need me: mla signs me in through my browser, and Codex hooks need me to grant trust. Tell me when you need me and I will do my part.
Both agents are supported today. Install both and each is governed by the same approved decisions.
Everything it installs is open source. Read the client at github.com/Meetless/mla (Apache-2.0).
It is beta software. Paste the prompt and you are in.
- Your agent installs mla and signs you in. A personal workspace is created for you automatically. No invite, no waiting on us.
- It moves fast: expect rough edges and the occasional breaking change while we are in beta.
- It is free while in beta, and we answer fast when you hit a snag.
Works with Claude Code and Codex
mla governs both major coding-agent CLIs through one neutral decision core. The loop is identical for each; only the wiring differs. These are siblings, not alternatives: install both and each agent is governed by the same approved decisions, because the decision logic lives in the core rather than in either connector.
Claude Code
Wired by the installer
- 1Installing mla configures Claude Code on this machine, automatically.
- 2Sign in with mla login.
- 3Bind the repo with mla activate.
Your turn: Complete the browser sign-in when your agent asks. That is the only step it cannot do for you.
OpenAI Codex
Wired by the installer too
- 1When Codex is installed and available on this machine, installing mla configures its MCP server and hooks automatically.
- 2Sign in with mla login, then bind the repo with mla activate.
- 3Run mla doctor to confirm both halves are live.
Your turn: The browser sign-in, plus one more: after restarting Codex, open /hooks once to review and trust the Meetless hooks. Until you do, Codex skips the hooks and governance is inactive.
Tested against Codex CLI 0.144.6. mla codex uninstall removes only the Meetless entries from $CODEX_HOME/hooks.json, leaving your own hooks and your Claude Code wiring intact.
→Prefer to run the commands yourself?
You do not have to. The prompt above lets your agent do all of this. But if you would rather run it by hand, or just want to see exactly what your agent will run, here are the same commands.
Install
Install mla on macOS or Linux. Pick one; all three install the same single binary, mla.
brew install --cask meetless/tap/mlanpm install -g @meetless/mlamla is a single binary with no required config. The curl and Homebrew paths need no Node at all.
Verify
To check that mla installed correctly, open a new terminal and run:
mla --versionIf this prints a version number, you're ready.
First commands
The install wires mla into Claude Code, and into Codex when Codex is present. Two conscious steps remain: sign in, then bind a repo to a workspace.
1. Sign in (opens your browser). We create your personal workspace automatically on first sign-in:
mla login2. Inside the repo you want to work in, bind it to a workspace:
mla activateThat is it. mla is now wired into Claude Code in that repo, and it starts grounding your agent in your team's decisions right away.
Wire up Codex
Installing mla already wires Codex when Codex is present, the same as Claude Code, so most people never run anything here. Codex has two independent halves: the MCP server (governed retrieval) and the hooks (grounding injection and the pre-execution gate). Run this only if mla was installed before Codex, so the auto-wire was skipped, or to re-assert the wiring. It writes both halves, $CODEX_HOME/config.toml and $CODEX_HOME/hooks.json, and is idempotent:
mla codex installThen run mla login and mla activate as above, and confirm both halves are live:
mla doctorHooks fail open until you trust them.
After restarting Codex, open /hooks once to review the MLA entries and grant trust. While the hooks are untrusted Codex silently skips them: governance is inactive and tool execution proceeds normally. No agent can grant this for you.
Enforcement is advisory by default. Every rule is clamped to warn until you raise the ceiling for a session with MEETLESS_ACTION_INTERCEPT_MAX=deny. Denied and warned attempts are captured as enforcement incidents and surfaced by mla enforcement --all.
Security and verification
Cautious about piping a script to your shell? Good. Read it first:
curl -fsSL https://meetless.ai/install.sh | lessPin a specific version:
curl -fsSL https://meetless.ai/install.sh | MLA_VERSION=0.1.0 shEvery release ships a SHA256SUMS file and a signed manifest. The installer verifies the download automatically and aborts on any mismatch.
Upgrade
Upgrade in place to the latest signed release:
mla upgradeOr upgrade with the method you installed with: re-run the install script, brew upgrade --cask meetless/tap/mla, or npm install -g @meetless/mla@latest. Your config is untouched either way.
Uninstall
Remove mla and its local footprint:
mla uninstallTo remove just the Codex wiring and leave everything else alone:
mla codex uninstallJoin the community
Get help, share what you are building, and help shape mla with other developers. I'm most active here, so ping me directly and you'll usually get a quick response.
Questions?
Hit a rough edge or not sure where to start? Email us → We answer fast during the beta.