Team chat · governed agents

Less slack.
More ship.

Team chat with AI agents that ask before they act. Permissions, audit logs, and a workspace that stays out of the way.

Create a workspace What's different

3-day free trial · Web + PWA · Open roadmap

#product14 people · 3 agents

Onboarding flow review14 replies · 2h

Marta: "Step 3 still feels heavy. Two options on the table for Friday."

Pricing page experiments8 replies · 4h

Agreed on three variants for next sprint. Owner: Devin. Summary pinned.

Stripe · new subscriptionbot · 9m

New Pro sub — €49/mo. Standup agent added a thank-you draft for review.

Proposed · awaiting approval

Why Tight

Chat hasn't kept up with how we work now.

Agents are doing real work. Your chat app should know what they did.

Chat stopped evolving

Channels, DMs, threads. The basics haven't changed in a decade — and the bills keep climbing.

AI got bolted on

Summary buttons. Assistants that hallucinate. Nothing that actually runs work on your behalf.

Bots run wild

Install a bot today and it can do anything, anywhere, with no audit. That doesn't scale.

Switching hurts

New chat app usually means moving the whole team on Monday. Tight lets you run both for as long as you need.

What's different

Three things we do that chat apps don't.

From chat to done

An agent watches your channels, spots the tasks nobody wrote down, and turns them into assigned work — with context intact.

Detects commitments in plain conversation
Assigns, sets due dates, tracks progress
One click to reject a false positive — the bot learns

Agents with guardrails

Every bot has a policy, an allowlist of channels, and an audit log. It proposes, you approve.

Per-channel allowlist, enforced server-side
Recommend-before-act by default
Every action logged and attributable

Skills you install, not bots you build

Pick a skill from the OpenClaw library, drop it into a channel, done. MCP under the hood so any model can run it.

Drop-in skills for standups, PR triage, and more
Bring your own model — Claude, OpenAI, or self-hosted
One-click migration from your existing Slack workspace

Task extractor

From a casual line to a shipped to-do.

A real example: Marta mentions something in passing. The agent proposes a task. Sam picks it up.

1 · a line in chat

#product

Marta09:18

@sam can you update the pricing page copy before Friday?

2 · bot proposes

◆

Task extractor09:18

Spotted a task: Update pricing page copy. Assign to @sam, due Fri?

3 · in Sam's inbox

Update pricing page copy

Due Fri·#product·from Marta

Skill library

Install a skill in one click.

Every skill ships with its policy preset. Allowlist the channels, decide action mode, done.

◆Task extractorInstalled

Watches channels, spots commitments, creates assigned tasks.

◇Standup runner

DMs the team, collects yesterday/today/blockers, posts a digest.

◈PR triage

Routes pull requests to reviewers and nudges stale ones.

◉On-call rotator

Reminds the on-call, swaps coverage, flags incidents.

◊Sentiment watcher

Surfaces channels trending negative so leads can step in.

◎Answer desk

Drafts replies from your docs when the answer's already written.

How it works

Three steps. Under five minutes.

Create a workspace

Chat

Channels, DMs, threads, reactions, mentions, search. Everything you'd expect.

Add a bot

Pick a BotAgent, allowlist some channels, watch it work. Approve every action until you trust it.

Skills · not bots you build

An agent spots the task before you ask.

Install a skill from the OpenClaw library. It watches the channels you allow, proposes work, and never acts without approval.

✓Sees only the channels you allow
✓Asks before acting
✓Logs everything
✓Replies when @-mentioned, silent otherwise
✓Any MCP-aware model can run it

Try it in your workspace