Bring your own Agent to MS Teams
microsoft.github.io56 points by umangsehgal93 9 hours ago
56 points by umangsehgal93 9 hours ago
I just want Teams to have a proper webhook API!
Why do I have to create a weird "workflow" to receive webhooks in a chat channel? Why does it have such complicated ownership permissions? Why is it automatically deleted if it doesn't see "enough" use? Why do I need to send the message in some weird "adaptive card" format - for which there are barely any ready-made libraries? Heck, why can't I create a new chat channel on my own, but need to invite three other people into it? Why does chat have a completely separate "channels" functionality which act more like a forum?
Please, stop trying to reinvent the wheel! Just do exactly what all your competitors are doing and I'd be so much better. My company is already stuck with Teams due to M365 licensing, there is no need to make it "unique".
Imagine when you receive millions of events per day -- the ones that arrive out of order, the days when delivery time goes up and up again, the days when oauth fails renewing keys... it's meant to be a lot of fun.
Compare to a sad websocket that just stays connected, you receive everything in order and you don't need an harness with tunnels every time you want to test something in dev.
Well, if you are stuck with M365, that may be the explanation. There's little market pressue on Teams.
Your prompts will go through so many companies, who all want to have a piece of that sweet human conversation.
Teams / IDE / Harnass / Router / Agent
> Your users live in Teams.
No, I assure you, we do not. Nobody wants to live in hell. Even at my current company, we are mandated solely to use Teams for meetings. But nobody uses it for chat.
Thankfully we have Slack to save us from that hell. Even if giphy is disabled because it might have R rated content whatever the hell that means.
>> Your users live in Teams.
> No, I assure you, we do not
This absolutism is tiring. I can assure you there are plenty of companies who use teams as their sole messaging platform. My company uses slack, teams, and zoom. Which one you use the most depends on the person or the team they are on. (Although zoom is being phased out to the chagrin of some folks).
Teams isn't without it's issues, it's annoyingly slow, and startup takes so long I have missed a meeting or two because of it. But, the conferencing experience is pretty decent, and I think you're just being contrarian for the sake of it.
> This absolutism is tiring.
What is really tiring is using Teams as a tool to embed other websites. Because that's what "apps" and most other functionalities are.
Teams is just a very slow web browser with the worst possible way to manage bookmarks and tabs.
You clearly don't have an organization fully onboarded to Teams, as you yourself describe. You use it as a conferencing tool. When you add all the other bloat Teams promotes it becomes a burden that makes me want to come back to sorting thousands of emails with outlook rules.
Fortunately, there are companies which still use IRC and other entirely-open standards such as SIP for collaborative communication.
Sounds nice. I've seen places that have essentially given up email in favor of teams for all communication.
IMO Only thing good about Teams is Emojis
Cant wait to look like an idiot after sending a heart to my boss while trying to select the hostname
Do you mean the way that Microsoft hijacked open-parenthesis, and you can't choose a different character to trigger an emoji call?
IMHO they are not. Using the shortcodes via () works okay-ish, but open the emoji menu is just awfully slow, it takes a solid 5-8 secs for it to show the emojis not recently used. It's absolute garbage.
Except for the fact where their emoji :shortcodes: are different from everyone elses, so if you're not careful you end up looking like a lunatic.
I know it’s NerdCool to hate on Teams, but I really don’t get it.
I’m in Linux with portal for teams (electron app)… and it’s fine. Like, we moved to it from slack and I just don’t see much of a difference. I type stuff, and people see it. I set up a call and people join, we talk.
What is it that is so broken, or is it just cool to hate?
> I type stuff, and people see it
I wish... I'm on Windows 11 Enterprise with corporate Teams, definitely 100% Microsoft approved combination and:
- on one channel my message never got sent, while on another channel it does work
- sometimes when I scrolled up the old messages doesn't show up, with a 'message removed by organization retention policy' text.. but those are messages from yesterday and sometimes when I restarted Teams it shows up again.. sometime it doesn't but when I opened web Teams it does show up
- sometime I can't connect to Teams for no reason, restarting Teams and computer doesn't help either, went to the IT helpdesk and they spent several minutes redoing what I did until they just googled it and delete the cookies or something
I also used slack and from my perspective its 100% reliable at delivering text messages
I use Linux and try to use Teams in Chrome for some client communications. It's been a frustrating experience to get myself logged in. There's a lot of confusion about my live account, microsoft account and several others that i'm not familiar with. I usually just run it without logging in now and just type my name.
Slack is definitely better but I have had issues with message delivery and things especially across the "native" desktop application and the mobile app.
Flaky Wifi connection? Maybe Slack just retries more, and Teams gives up easily?
People have different aggravation thresholds. Some will complain at small things very quickly, while others will let the resentment build up until one day they snap. In relationships the former group tends to fare best in the long term. Since Teams is evidently not perfect, I can only assume you are in the latter group.
Try being a consultant who needs to have access to 2+ Teams accounts simultaneously.
On Slack I can just press command - <number> and I'm switching between client Slack accounts.
On Teams? Nope. Nope. Nope.
I suspect those who honestly think Teams is fine have never used anything better.
this is vibe coded :D. the drop down for Python not working.
I remember when we started working on our teams bot in 2023, Microsoft announced that they will stop supporting Python for the teams sdk, which felt super short sighted. Eventually they silently picked it up again and the old sdk never stopped working.
Kudos for turning it around. I believe it'll last this time as AI agents in communication tools are something a lot of people want. and unlike pre LLM chatbots and agents, they are so much more useful now.
Interesting to see the SDK making chat/bot integration simpler, but there's a whole other dimension to Teams integration that this doesn't touch: telephony.
The company I'm at has been been building call analytics for Teams since 2021 (QueueMetrics Live). A long-time customer migrated their entire phone system to Teams during the pandemic and asked my boss if we could follow them there. We said "in principle, yes": the Graph API exposes telephony events, so it seemed doable.
It was doable. It was also far harder than anybody expected. The Graph API gives you raw call records, but it doesn't model concepts like "wait time in queue" or "lost call" or "failed agent attempt", and whet it does, nowhere it written HOW. You have to reconstruct those from sequences of low-level events. We ended up building the whole processing pipeline in Elixir because we needed to handle the real-time stream reliably at scale. After a long beta and a few million calls processed, we got to GA. I was dabbling in Elixir at the time, I put it on my CV, and it was noted. In a sense, I owe Teams my current job (though I ended up on a different team, so I'm mostly referring coffee-mug lore here).
We set out to track queues and auto-attendants (that's our bread and butter from the Asterisk world), but we discovered we were seeing everything — inbound, outbound, Teams-to-Teams, even calls with other companies with their internal ids. You can get a complete picture of someone's telephony activity regardless of whether they work in a contact center or just use Teams as their phone. Most of the boring config (names, groups, codes) comes straight from Graph, which is nice.
Like it or not, a lot of enterprises are quietly moving their entire communications (including telephony) to Teams. When they do, they lose the monitoring and analytics they had with their old PBX. That's a real gap, and the Graph API — despite its limitations — gives you something to fill it. But I have a feeling that "3 lines of code" won't cut it. :-)
Can all the king's agents and all the king's men stop Teams from frequent silent failures to deliver a short message in Year of the Lord 2026?
Did they call this Copilot yet?
Why is there no Teams CLI?
There's a libpurple plugin, so you can in theory use Teams chat from Pidgin, finch, or other clients. I tried it last year and it didn't work very well.
There is an API:
https://learn.microsoft.com/en-us/graph/use-the-api
I've been using it recently and aside from some weirdness around permissions, it's decent.
If you need a CLI wrapper I'm sure an agent could make it for you in an hour or so.
Was this written by an agent? It’s giving me a pain by the second paragraph.
I also found it truly brutal to read and got fed around the same spot. I hate this style so much.
> Teams is where most enterprise work happens
It's nice that they start the post with some sarcasm
If only it could fix the lag with Mac screen sharing in Teams.
When I were a nipper lag/latency over 30ms was considered a bit crap for voice unless satellite links were involved. That's for circuit switched networks. Human conversation works best with a sub 25ms latency and you will start to notice lag at 30ms.
Nowadays with all our massively more powerful links (Gb vs Kb) but packet switched, we often end up resorting to a form of half duplex radio protocol. That's just voice, let alone video.
That's what you get when you abrogate your comms to a hyper scaler that will never scale to the point of what you would like because it will damage profits upstream.
Whilst your end will be a phone or laptop or whatever - with gobs of capacity, the hyper scaler bit will be woefully under powered for your call but just enough to keep comms going and your subscription dumping cash into the coffers.
You end up re-inventing how to talk to someone over a satellite link in the 1970-80s ... in 2026! I (UK, 55 y/o) can clearly remember my parents telling me how to talk to great aunt Maye in Australia on the blower. Nowadays we have the internet to packet switch instead of circuit switch which is generally capable of ~10-50ms latency nearly anywhere, where mostly copper is involved. However call quality seems to be shit!
How much did a phone call cost per minute from UK to Australia in the 70's and 80's ?
The modern experience is not perfect but audio is usually far better than the 70's and 80's, video is often an option, and in general the experience is orders of magnitude cheaper.
PCMU and PCMA voice frames are 20ms or 40ms, and no one is running with no jitter buffer, so your 30ms number doesn't make sense.
Even circuit switched networks are not often below 30ms, to hit that you'd need to make a local phone call on a fully analog circuit.
How many levels of agents are here. Agents riding code by agents in a system driven by agents vibed by one lonely engineer in Redmond?
Introducing Microsoft Teams Turducken 2026 (Enterprise AI Agent Edition) now with 17 layers
https://www.npr.org/sections/thesalt/2014/11/21/365509503/th...
I've been having a bit of success with using Matrix with OpenClaw. I've been playing with OpenClaw because our company is looking to help people get started with AI workflows. So, I've been experimenting a lot with different tools and technologies in this space.
I started like everybody else trying to do simple things via Whatsapp and Slack. Terrible experience. Both systems are pretty locked down and very inflexible and the OpenClaw integration is very limited. You need a phone number for each user in Whatsapp and if you don't have an extra one, you basically end up chatting with yourself. With Slack you end up in permission hell. And their UI for editing permissions is terrible. I never quite managed to make this work.
With Matrix, the experience massively improved. I self host it and I got codex to generate me a working setup that has been running on a cheap vm for a few weeks now. Backups and everything. Hooking that up to OpenClaw was pretty straightforward. You just get a user/password and it gets an access token and then you are good to go to do whatever via the REST API and CLI. So, the OpenClaw integration worked on the first try for me.
Then I had a nice idea and took this to the next level: I gave codex access to my vms with OpenClaw and Matrix stack (Synapse & Element) and I made it create an OpenClaw Admin agent for me, with its own matrix admin user. And then I "taught" it to create more rooms, agents, and bot users for me.
Now I can prompt "create an agent called Foo" via matrix and it will invite me and the rest of our team to the new room. And a minute later I can be chatting to my new agent via a freshly created bot user and model of my choice. Super simple.
We actually ditched Slack a week into this experiment. Because obviously I got the Admin agent to also invite the rest of the team to these rooms and everybody started engaging with agents. We made the admin agent recreate our dozens of slack channels we had as rooms from a screenshot. Migrating content from Slack is a bit meh so we just just skipped that. But Slack in general is a bit meh these days so we don't really miss it.
MS seems to be catching on to this with Teams. I still don't like it. But providing an SDK for it is the smart move. It's essential to enable doing stuff like this.
Is OpenClaw useful? It can be. But you need to solve a lot of plumbing issues like this. Getting the basic plumbing out of the way for setting up new agents with communication channels is a nice first step. We are currently experimenting with all sorts of simple use cases currently. They all require setting up a dedicated agent and having OpenClaw do that for us is both nice as a demo and something we now use a lot.
[dead]