A shared agent for your team

A lot of work should not depend on one person.

Loopable is your team’s own agent. Work comes in from Slack, Lark, GitHub, or a schedule. Codex, Claude Code, or Cursor Agent does it on a machine you own, and the answer goes back where it was asked.

#oncall Slack bot

Maya

@loopable checkout 5xx since 4pm?

Office Mac Idle

Codex

Waiting

#oncall Slack bot

Loopable

Cause: redis timeout on cart

npm install -g loopable-cli && loopable start What you need →

The problem

Team work still waits on one person.

01
Someone has to be there

A review request, a question in #oncall, a morning check. Someone copies it into their own Codex. If they are busy or away, it stops.
02
Know-how lives on laptops

Everyone has their own prompts, skills, and config. The answer depends on who asked. Nothing is reused, and nothing is recorded.
03
Hosted agents can’t reach in

A cloud sandbox can’t see your internal network, logs, or services. You get one vendor, and usually one chat tool.

Shared

One agent the whole team hands work to.

The team leaves Loopable running. A signal comes in, a loop matches, the agent runs, and the answer is written back.

No one has to be there

A loop runs when a review is requested, when a chat bot is mentioned, or when the clock says so.
One way to do each job

A loop says what to watch, what to ask, and where to answer. Anyone on the team can read it and change it.
Know-how in the repo

A loop works in your repository. The AGENTS.md and skills there are what the agent knows.
Every run on the record

The inbox shows what came in, what the agent did, and what it wrote back.

Yours

It runs on your machine. You decide what it can do.

The agent runs on a computer the team owns. A Mac in the office is enough.

Reach
Whatever that machine can reach: the internal network, logs, databases, internal services.

Hosted: an allowlist out of a cloud sandbox.
Extend
Any MCP server, CLI, or internal login you set up on it.

Hosted: the connectors the platform ships.
Choose
Codex, Claude Code, or Cursor Agent, and the model you want.

Hosted: one vendor.
Keep local
Code, logs, and credentials stay in-house. It uses the CLI plan you already pay for.

Hosted: your work runs in their cloud.

What it does today

Loops you can turn on now.

Any trigger Custom loop: your own prompt Anywhere, or Inbox only
Schedule Do something on a schedule Inbox, or a Slack, Lark, or GitHub thread
GitHub Review pull requests A review, as a comment
GitHub Reply to issues A comment on the issue
Slack bot Do what I ask the Slack bot A reply in the thread
Lark bot Do what I ask the Lark bot A reply in the thread
WeChat Do what I ask the WeChat bot A reply in the chat
Gmail When an email arrives A reply in the same thread, or anywhere

How it works

Watch. Run. Write back.

Slack bot Lark bot GitHub Schedule

Loopable App + Dispatcher

Watch
Polls Slack, Lark, and GitHub.

No public URL needed.
Match
Pick the loop.

Each signal runs once.
Queue
Hold the job.

A runner pulls it.

Runner Your machine

Pulls the job and runs the agent there, with that machine’s network, tools, and logins.

Codex Claude Code Cursor Agent

Slack thread Lark thread Review or comment

Run it

One computer that stays on.

A Mac in the office is enough. It needs Node 22+ and Codex, Claude Code, or Cursor Agent signed in.

1 · Start

npm install -g loopable-cli
loopable start

2 · Write a loop Open http://127.0.0.1:4321 on that machine. Connect GitHub, a Slack bot, or a Lark bot, and pick a loop.

3 · Join a runner

loopable runner \
  --url http://<LAN-IP>:4321 \
  --token <from Runners>

A lot of work should not depend on one person.

Team work still waits on one person.

Someone has to be there

Know-how lives on laptops

Hosted agents can’t reach in

One agent the whole team hands work to.

No one has to be there

One way to do each job

Know-how in the repo

Every run on the record

It runs on your machine. You decide what it can do.

Loops you can turn on now.

Watch. Run. Write back.

One computer that stays on.