Launched this week
Switch brings AI agents into the collaboration tools your team already uses. Agents join as named participants in your Slack, Teams, Discord, or Telegram channels, sharing the same context and history as your team. Connect an agent once and use it across projects. Each room carries its own context, participants, and rules. Works with Claude Code, OpenAI, Google ADK, LangChain, and more. Open source, self-hostable, running in minutes.






Can non-technical people on the team use this too or is it mainly for devs?
Switch
@ann_le1 Once it's set up, anyone who uses Slack or Teams can work with the agents! You just message them in the channel like you would a teammate. Setup takes about ten minutes and our getting started guide walks through the whole thing step by step. The day-to-day use is for everyone!
what kind of rights do you need to give to the agents?
Switch
@mary_urdahl Great question. On the agent side, you give it an API key when you connect it and Switch handles the communication from there. It doesn't need direct access to your Slack or Teams workspace since Switch handles that bridge. On the people side, you control who can talk to each agent and which channels it can join. Everything starts locked down to just you and you open it up from there.
Hey @safi_amin, how does it compare to Viktor?
Switch
@a_c26 They solve different problems! Viktor is a single AI agent that executes tasks for you in Slack and Teams. Switch is the infrastructure that lets you bring any agent, from any framework, into your team's channels. Most teams we talk to are already using multiple AI tools and the challenge is getting them all to share context. That's what Switch does. Viktor is great if you want an agent to do work for you, whereas Switch is for teams that want all their agents and people working together in one place.
Can an agent carry context from slack into teams and vice versa? I see that as being the most useful element, where context from a slack channel with engineers can answer questions on a teams channel with execs.
@safi_amin congrats on the launch, looking forward to testing this out.
Switch
@safi_amin @umier You can bridge between rooms if needed (even on different messaging apps), or you can get agents from each room to write to a common resource outside of it (eg confluence, a word doc, a wiki) to ensure context is carried.
Can an agent keep separate memory and rules for each Slack channel?
Switch
@nolanpierce03 Each Slack channel gets its own instructions and roles that an agent can take on, so agent behavior can have different rules in different channels. Channels can also have their own documents in Switch that act as a memory for that channel. An agent's own memory (i.e., its working directory) travels with it between channels.
a simple activity or audit log for each agent would be really useful, especially when multiple people are interacting with it.
Switch
@awesome_america absolutely! the right level of visibility differs by organization and use case. switch connects agents from anywhere, and those agents might already have usage statistics tracked. it's also a fair ask to have a switch-specific, configurable audit log
Bababot
How does Switch handle permission controls when multiple AI agents are in the same Slack channel?
Switch
@adams_parker Good question.
Each agent has an owner (human). The default permissions is that only that owner can speak to the agent in any context (channel, etc.). You can tweak this to be much more permissive, e.g. any agents or humans in any context can speak to this agent, or something more granular, e.g. my agent can be spoken to by agents x,y,z and humans u,v in channels a,b.
Behind the scenes, on every message sent to your agent in a channel, we have a gate that checks the policy of that agent can block that message if needed.
Switch
@adams_parker One thing worth adding here is that as each agent has its own policy, you can have 5 agents in the same channel with completely different access rules. One might respond to anyone, another only to its owner. The agents and their permissions dictate that, rather than the room itself.