Switch - Bring any AI agent into Slack, Teams & Discord
by•
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.
Replies
Having one environment that keeps the organizational context and workflows consistent while allowing for the other tools (Claude, Slack, etc) is a big unlock for making human x agentic collab work durably. Will be testing this out soon
@luca_ferrara2 Yep 100% That's the vision -> provide the underlying infrastructure and framework that allows for true human x agentic collaboration.
@luca_ferrara2 Would love to hear how it goes!
Hey @safi_amin, how does it compare to Viktor?
@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.
Really like the idea of agents becoming part of the team’s existing conversations instead of another tool people have to switch to. Congrats on the launch, Safi and team!
If I built my own harness, how would I bridge it to Teams with Switch? (I did build one ;))
That looks pretty cool. Congrats to the build and launch :) Will give it a try. Isnt it quite token intensive if your agents are always following the conversation?
I noticed we sometimes have discussion which then get compressed into a short clear working document for the agents to run it. Lots of answeres already cleared beforehand. I do have the feeling though the vision of switch is bigger and i will need to grow my horizonts :)
Really cool product - I like it
The idea that you can bring any AI agent into the tools your team already uses without rebuilding anything is so practical.
@maria_anosova Awesome to hear that Maria! Yes, we really wanted to keep this low friction for users
Congrats on #1 today. The "named participant with shared context" framing is what stands out - most Slack bots reset context every message so the agent never actually feels like it's in the room. Curious how conflicting instructions get resolved if two humans in the same channel give the agent contradictory asks in the same thread.
@galdayan thank you! the persistent context definitely opens things up. When it comes to conflicting instructions, the agent just has a single context window so it will see both instructions and the LLM "brain" will reason through what to do. Switch itself doesn't arbitrate. Of course, it may be that only one of these humans is allowed to instruct the agent, in which case the other human's message doesn't reach the agent as an instruction at all.
@galdayan Thanks Gal! Appreciate you checking out Switch and leaving a review. The feedback on the docs is super helpful.
When an agent connector drops mid-thread, does Switch retry delivery or hand the room back to a human? Congrats on the launch btw!
@colegawin Hi there! Switch does not need retry the delivery. Upon reconnecting, the client on the agent side will tell the switch server what is the last message he has seen, and then switch will make sure to send everything that was missed to the client and agent
@lamaudruz Got it, makes a lot of sense :)
@colegawin The nice part is this happens invisibly! From the team's perspective in Slack or Discord, the agent just picks back up where it left off. Everything stays in sync automatically and the team doesn't have to do anything.
Love the approach of treating agents as real channel participants. The open protocol is the right call here, really smart way to solve the last mile problem for AI agents. Congrats on the launch!
@ana_robakidze Thank you! Last mile is a good way to put it. The agents were already capable, the gap was getting them to where the work happens with the rest of the team.
@lamaudruz wrote a great blog post outlining how we got here! https://www.flintai.dev/blog-posts/our-team-is-now-composed-of-5-devs-and-45-agents-all-in-slack