LAINTAS CLI · THE AI AGENT IN YOUR TERMINAL

Your shell, as usual.
Plus an agent.

Laintas CLI is an autonomous agent that lives in your shell. Commands like ls, git and vim still run straight in a real PTY, never through the model. Describe a task in a sentence and the agent reads code, runs commands and edits files — every step in front of you, every risky one waiting for your yes.

curl -fsSL https://cli.laintas.com/install.sh | bash
A real session, recorded as it ranRecorded on a real machine: the agent reads the whole project, runs the tests, verifies the output itself, and only then writes a file. Nothing is staged, and the model pauses are not cut.
Free
Free runtime, source you can download and audit
Linux · macOS · Windows
x86_64 / arm64, Windows 10 2004+ / 11
One account
Shared with Helpwo and the plugin market

01 · WHY IT EXISTS

The agent closest
to your files.

An AI in a chat window can only advise. Copying, pasting, running, and relaying the error back is still your job. Laintas CLI stands in your working directory instead.

Chat windowLaintas CLI
ContextThe snippets you pasted inThe whole workspace, terminal state and project memory
ExecutionHands you a command to runRuns it, reads the full output, decides the next step
On failureYou relay the error backIt sees the same screen of errors you do
Long tasksClose the tab and it is gonePlans, events and sessions persist; /resume picks up

Real terminal

Direct commands bypass the model; interactive programs, long jobs and named sub-terminals keep native shell behavior.

Parallel work

Sub-agents split the job: HWO coordinates live collaboration, HWG compiles dependencies into resumable graphs.

Layered authorization

Modes, roles, phases, global policy, trust and approvals narrow what a tool may do — enforced by the runtime.

One tool surface

Built-in tools, Skills, MCP servers and plugin-market extensions share one registry and one authorization pipeline.

02 · THE LIFE OF A REQUEST

One request in.
Every step visible.

Take “fix the failing test”. The path the runtime actually walks:

  1. 01Route$ pytest → PTY
  2. 02Assemble contextmode: act · scope: project
  3. 03Call toolsread · shell.exec · edit
  4. 04Pass the gateallow · ask · deny
  5. 05Run and recordtrace: on

SECURITY BOUNDARY

Parallel,
never out of hand.

Prompts express intent; the boundary lives in the runtime. No amount of model persuasion gets past a policy that already said no.

Allow

  • Read project files
  • Run the tests
  • git status / diff

Ask

  • Write or delete files
  • git push
  • Read ~/.ssh and other credentials

Deny

  • Send credentials off the machine
  • Tools the policy disables
  • Sub-agent calls beyond their role

Which tier an action lands in depends on the active mode, project policy and organization policy.

03 · PRICING

Install first.
Decide how much later.

Laintas CLI itself is free. Model calls are pay-as-you-go or by membership, and every plan's allowance is shared across every Laintas product.

Full pricing and model tiers

04 · DOWNLOAD

Now,
hand it to your terminal.

The one-line installer is recommended. You can also grab a package directly, or audit and extend the source.

curl -fsSL https://cli.laintas.com/install.sh | bash
  1. 01Run the installer

    It detects your architecture and installs laintas-cli

  2. 02Start and sign in

    Type laintas-cli and sign in to your Laintas account

  3. 03Give it a first task

    Try “figure out how to run this project”

Packages

SHA256 checksums for every file

Linux needs a 64-bit glibc system. Mac supports Apple Silicon and Intel without Helpwo Kernel. laintas-cli-windows needs Windows 10/11 x86_64 and WebView2; the installer can download WebView2. WSL is not needed.

Helpwo windows: the desktop edition for Windows 10/11 (64-bit), using the same Laintas account. Install and useHelpwo windows SHA-256 checksum