Twin for Slack
Your AI agents, working where your team already talks.
Twin lets anyone build autonomous AI agents that run real business workflows — researching leads, updating your CRM, compiling reports. The Twin Slack app brings those agents into Slack: they post briefings to your channels, pick up tasks when you mention them, and answer in DMs, so work gets delegated and delivered without leaving the conversation.
What Twin does in Slack
Most teams lose hours copying information between tools and chasing status updates. Twin closes that gap: agents do the work in your connected tools and use Slack as the place where it's requested, reviewed, and delivered.
Scheduled briefings and digests
Agents compile data from your other tools and post structured summaries to the channels you choose — daily pipeline reports, renewal queues, error alerts — at the times you set.
Delegate work with @Twin
Mention @Twin in a channel to hand off a task. It replies in the thread, does the work in your connected tools, and reports back exactly what it did.
Direct messages with your agents
Message Twin one-on-one from its Messages tab — or with a slash command — to kick off a run, check on a workflow, or ask for results. Share files as inputs and get reports and exports delivered back.
Channel monitoring and triage
Invite Twin to a channel and an agent can classify incoming messages — support requests, commands, reports — and route each one to the right place.
Connected to your whole stack
Slack rarely works alone. A single agent can read a spreadsheet, update your CRM, send an email, and post the summary to Slack — no glue code required.
Humans stay in the loop
Before consequential actions — sending an email, updating a record — Twin asks first. Approve or hold directly from the Slack message.
Installing Twin in your workspace
Installation takes a couple of minutes and uses Slack's standard OAuth flow. Nothing is installed until you've seen and approved exactly what the app can access.
- 1
Create or sign in to your Twin workspace
Twin is free to start. If you arrive from the Slack Marketplace listing without an account, you'll be guided to create one or connect an existing one before installation continues.
- 2
Connect Slack from your Twin workspace
Choose Slack in Twin's integrations. You'll be redirected to Slack's standard OAuth consent screen, which lists every permission the app requests before you approve anything.
- 3
Approve the installation
Once you approve, Twin confirms the installation and walks you through your first workflow, so you go from install to a working agent in minutes.
- 4
Put @Twin where you want it to work
Private channels require an explicit invite, and you can remove Twin at any time to cut off access. For public channels, Twin joins only the specific channels you configure an agent to post in or monitor — never your workspace at large.
Changed your mind? Uninstall Twin from your Slack workspace's app management page or disconnect Slack inside Twin — either way, the app's access tokens are revoked. Need a hand? Support is one email away.
The permissions Twin requests, and why
Twin follows the principle of least privilege: it requests bot permissions only — no user-token scopes that act as you, no admin scopes, no legacy scopes, and no workspace-wide message access. Each permission below maps to something you can see the app do.
chat:write Post messages and thread replies as Twin in the conversations it's part of.
chat:write.public Deliver an agent's output to any public channel you select as a destination, without the app having to be invited there first.
chat:write.customize Post with a per-agent name and avatar, so every message is clearly attributed to the specific agent that produced it rather than a generic bot.
app_mentions:read See messages that @mention Twin so it can pick up tasks delegated to it in a channel.
channels:read, channels:join List your public channels so you can pick where an agent posts or monitors, and join the specific channels you point it at. Twin never joins channels you haven't configured.
channels:history Read messages in the public channels Twin is a member of — this powers monitoring, triage, and replying with full thread context. It is never workspace-wide access.
groups:read, groups:history Work in private channels Twin has been explicitly invited to: see that the channel exists and read its messages. Private channels Twin isn't in remain invisible to it.
im:read, im:history, im:write Run the one-on-one conversation in Twin's Messages tab: read the messages you send it and reply to them.
mpim:read, mpim:history Take part in group direct messages that include Twin, under the same rules as one-on-one DMs.
commands Register Twin's slash commands so you can trigger agents straight from the message box.
files:read, files:write Receive files you share with Twin as workflow inputs, and upload generated outputs — reports, spreadsheets, exports — back into the conversation.
pins:read, pins:write Keep important agent output pinned — for example an always-current briefing — and read existing pins for channel context.
reactions:read, reactions:write Use emoji reactions as lightweight signals — acknowledge a request, mark a thread as handled — and read reactions left on Twin's messages.
users:read, users:read.email Resolve member names for clear attribution, and match Slack members to their Twin accounts by email. These addresses are never used for marketing and are never contacted without explicit consent.
team:read Read basic workspace details — name and icon — to label the connection inside your Twin workspace.
incoming-webhook Create a webhook for the single channel you select during installation, used as the default delivery channel for that connection's notifications.
This list mirrors the consent screen you see during installation. If a future feature ever requires a new permission, Slack will ask you to re-approve the app — access is never expanded silently.
How we handle your Slack data
Twin Labs SAS, based in Paris, processes Slack data solely to run the workflows you configure, under our privacy policy and Data Processing Agreement. These are the commitments that govern the Slack app:
No training on your Slack data
Twin never uses Slack data to train large language models — neither our own nor our providers'. Our agreements with model providers such as OpenAI prohibit them from using your data to improve their models.
No message export or archive
Twin processes messages to run the workflows you configure. It does not export, back up, or build an archive of your message history, and it does not circumvent the history limits of Slack's free plan.
Slack's privacy model is respected
Agents read only the conversations they're part of — public channels you've configured and private channels Twin was explicitly invited to — and never surface private channel content, names, or membership to people who couldn't see them in Slack.
Short retention, real deletion
Workflow artifacts that contain Slack data are deleted within 30 days of a run, or sooner. When you uninstall the app or disconnect Slack from Twin, tokens are revoked and cached Slack data is removed within 30 days.
Credentials are locked down
Slack tokens are encrypted at rest in isolated secrets management, never logged or exposed client-side, and accessible only to a restricted group of Twin engineers for operational reasons.
Your data, your call
You can request access to or deletion of your data at any time by emailing support@twin.so. Full details are in our privacy policy.
AI transparency
Twin agents are powered by large language models. Like all LLM-based systems, the summaries, drafts, classifications, and other outputs they produce can occasionally be inaccurate or incomplete — review agent output before relying on it for important decisions.
Twin is built so that doesn't become a problem: agents show what they're doing while they work, report what they did when they finish, and ask for explicit confirmation before consequential actions. An agent will never silently join channels, message people who aren't expecting it, or make significant decisions without a human in the loop.
Details on the models we use, data retention, and data residency are available in our privacy policy and Trust Center, and we're happy to answer specific questions at support@twin.so.
Security
The Slack app follows Slack's security best practices for the Marketplace, on top of Twin's platform-wide security program.
Encrypted in transit
All traffic between Slack, Twin, and your browser is encrypted with TLS 1.2 or greater.
Standard OAuth 2.0 installation
Installation uses Slack's OAuth 2.0 flow with the state parameter to prevent forgery attacks. Granted tokens are scoped to the permissions listed above and revoked on uninstall.
Every request is verified
Every payload Twin receives from Slack — messages, mentions, interactions — is authenticated with Slack's request signing before it is processed. Unverified requests are rejected.
Independent security posture
Our security program, sub-processor list, and compliance documentation are published in the Twin Trust Center.
Found a vulnerability or suspect a security issue involving the Twin Slack app? Email support@twin.so and we'll respond as a priority. See also our Trust Center.
Support, without hoops
Email us at support@twin.so — no account or sign-up required. We respond to every support request within 2 business days, and usually much faster.
Bring your agents into Slack
Describe a workflow in plain English, connect Slack, and your first agent is reporting to your team's channel in minutes.
Get started with Twin