Skip to content

Developer tools

Installing agents in Slack skips the custom app

3 min read

Written by an AI model from the linked sources and checked against them by a second model. How we work

Image: thenewstack.io

Slack on August 20 launched Add to Slack. You start an install inside the tool where you built an agent, and the agent comes out the other end as a working Slack app in your workspace, with no custom integration code. Ten outside platforms support it at launch.

The partner tool does the setup that used to fall to the developer: OAuth, app configuration and permission scoping, Slack says in The New Stack’s report on the launch. Installation and agent creation both begin on the partner’s side. Before this, connecting an agent meant clicking through a series of Slack settings and turning on permissions by hand.

Doing that by hand meant registering an app in Slack’s developer console, getting a client ID and secret, setting redirect URLs and choosing scopes. Scopes are the list of actions the app may take, and the user sees and approves them at install. After approval, Slack sends back an authorization code, and the app’s server trades it for an access token. The partner now runs that whole flow.

The agent itself never moves into Slack. The runtime stays with the provider, or with tools like NanoClaw it runs on the user’s own machine or cloud account. Slack holds only the app. That app can receive messages and mentions and join channels, and it shows up in Slack’s App browser once installed.

Admins keep their existing controls. Installed agents inherit the workspace’s data boundaries and permission controls, Slack says, and the workspace’s app-approval policy still applies. What each agent can actually reach depends on two things: the OAuth scopes its app was granted and the channels it belongs to.

The launch partners are Hyperagent, LangChain, Lovable, n8n, NanoClaw, OpenAI, Runlayer, Skydive, Superhuman and Vercel. That list covers point-and-click builders as well as developer frameworks, and all of them use the same installation path.

Slack shipped a second agent feature the same day, Slack Code, and it does a different job. Slack Code creates temporary project channels where supported coding agents from its partners post plans, diffs, pull request details and live previews, and the channel is archived when the work is done. Add to Slack installs an app. Slack Code organizes what a small group of coding agents does after they are installed.

NanoClaw shows what the new path looks like for an open source framework. Until now, connecting it meant creating a custom Slack app and then setting up its configuration, scopes and authentication yourself. Under Add to Slack, NanoClaw’s Marketplace app acts as a manager app. It has permission to create a separate managed Slack app for each NanoClaw agent.

In a demo for The New Stack, NanoClaw creator Gavriel Cohen asked one NanoClaw agent to put together a team based on characters from The Office. “I say create a team for me of agents, and it creates all these agents and tags them. They all pop in, and then we start chatting over here,” Cohen said.

Each bot appeared under its own name, and Cohen also put them in one room with a shared Canvas. Behind this, NanoClaw opens a separate Slack WebSocket connection for each agent, and its host passes each message to the right agent in Slack.

“Each agent has its own bot. So it creates some new things. It means you can tag them natively,” Cohen said.

No comments yet. Be the first.

Join the conversation

Sign in to leave a comment. It appears after moderation.