I guess this makes sense for now. You can build integrations leveraging the user's personal access credentials. Later, once Claude takes over the world, they can move sessions to live in their own walled garden.
It reminds me that I don't really like Anthropic as a company, I just like Claude as a model a lot. It just feels more capable and personable than the others. I wonder if / when OpenAI et al. will be able to replicate it.
For now, I basically have no choice but to use the walled garden but I do hope Anthropic is not completely compromising their core mission of actually making the model better rather than following these public bandwagons.
Then again most of these probably take them like a day to develop through a junior dev talking to Claude Opus 5 or some shit lol (and to be fair, it shows). I don't know.
Claude Code daemon mode in background when?
My use case is that I have a separate system that provides human approvals for what my agent can do. Right now, I've had to resort to long-polling to give a halfway decent user experience. But webhooks are clearly the right solution. Curious to see how it ends up being exposed outside of these initial integrations.
However, once remote capabilities are added to any software, it is virtually guaranteed that they will eventually be exploited as backdoors.
This means enterprise security solutions will need to develop the capability to distinguish between legitimate and illegitimate Claude Code instances.
I presumed Claude would then be able to clone repos, make commits, update the code in its container and then write it back to github.
Instead, the github connector does ..... nothing it all. It's very weird.
What I wanted to build is a way for Claude Code to automatically receive reviews and CP failures from a Github PR and automatically revise code and respond to comments. It looks like with a custom Github PR channel I can get very close to this, although I do wish that a channel can be opened in a running session instead of having to create a new one. Hopefully they add that soon.
Basically the Claude CLI is the operating system is the product vibe I get right now.
Couldn't you have multiple sessions using different plugins or whatever?
Plus it gives a little ASCII dog to Claude Code terminal.
The ability to spawn independent CLI is awesome. No brainer they would add eventually between the great threaded functionality it brings and is essentially a more controlled version of OpenClaw IMO
Never had this problem with Claude tho. Must be something environment-specific.
Pushing events into an already running session feels like a step toward decoupling execution from transport state.
I've been thinking along similar lines — where a session continues to exist independently, and transports are just interchangeable carriers attached to it.
With this change, it looks like an officially sanctioned version of *claws. Connecting to whatever "channels" you want via MCP.
Architecturally it's a little different, most *claws would call the Agent SDK from some orchestrator, but with claude channels the claude code binary starts the MCP server used to communicate with the channel. So it's a full inversion of control where Claude code is the driver, instead of your orchestrator code.
I updated my nanoclaw fork to start the claude code binary in a docker container on PID 1, and you can read the docker logs straight from claude code stdout, but with comms directly to/from your channel of choice. It's pretty neat.
I tried getting it to use the `gh` CLI to do so but it either doesn't have the right permissions on its token or the requests are being intercepted and filtered by the sandbox it's in. I eventually dumped all the comments as JSON from my desktop and pasted it in at which point it handled them fine so it's certainly capable of working that way.
Maybe there should be a Claude code that facilitates others that is connected. Like sub agents but can "choose what to do" on permissions check.
Or some other means to listen for permissions check
It runs Claude in docker containers, listens for webhooks to see comments and CI status.
The rest of us were able to implement things like push a long time ago, but because Claude Code and Codex stubbed those things out, we couldn't really use them for 'most agent users'.
In fairness to OpenAI, they have been generous in allowing for example OpenCode to sign in with your ChatGPT subscription – so you _could_ build a more powerful agent (which OpenCode is... not) – but unfortunately GPTs' instruction following just isn't up to snuff yet. Hopefully they pre-train something amazing this year!
The pattern that works well: have the MCP server query sources in parallel, then use channels to stream results as they arrive rather than waiting for all sources to complete. For market research, this means the agent can start analyzing Wikipedia and Google News data while arXiv and GitHub queries are still running.
One thing I'd love to see is channel multiplexing — being able to label different data streams so the agent can prioritize which channel to process first.
Pi already has 700+ third-party packages [2] for various purposes of various quality. But it doesn't matter, since creating a new working Pi extension to suit your needs is just a prompt away, and you don't even have to restart your coding session.
[1] Pi Coding Agent https://pi.dev [2] https://www.npmjs.com/package/@e9n/pi-channels [3] https://pi.dev/packages
Telegram is 'bot friendly' since the beginning, gaining a lot of users with crypto boom a decade ago with coin drops and things like that, so it is very good to develop for, but I have your initial sentiment first - shame this hasn't launched with tools people actually use for work.
And no, Discord is not used for that either.
It's just as easy to set up as ntfy.sh, except that it doesn't break every other week on iOS.
iMessage is proprietary. WhatsApp charges you. Unofficial APIs exist, sure, but not my cup of tea.
Then you have Discord or Slack, which are pretty heavyweight when all you want is a simple chat interface.
Telegram makes it SO easy. Bots are first class resources on Telegram and they make them so easy to use.
Ctrl-Z $ bg
Or run it in tmux so you can pull it up on demand and have it open at startup.
I think CC does have “remote control” now which I think would work similar, but it’s Max only right now
"i enabled github connector can you see it?"
Answer: "I don't see a GitHub connector in the available integrations. The search only returned a Microsoft Learn connector (not connected). It's possible the connector hasn't fully activated yet, or it may not be available in your current setup. Could you double-check in Settings → Integrations that it shows as connected?"
Multiple such checks and re-setups do nothing.
I wrote it originally because I wanted my openclaw install to talk to my assistant's openclaw, and my openclaws that were local at different houses.
It's morphed a lot since then, and is close to being super useful -- it allows group chat, and is close to having a realistic API call on threshold vote gateway system built in.
That stuff is built to support Corpo's main business model which is providing real world asset and governance access to agents.
So, for example, I think agents might like to vote on sending a wire transfer by approving a specific mercury bank API call.
I could go on. You can also use it to remotely chat to an agent across firewalls - it's pull / poll only.
And if anyone is interested, I made an HN Group chat: https://chat.corpo.llc/?invite=p2F2AWR0eXBlZmRpcmVjdGVzdWl0Z...
Then I joined a group from a bigger city where I commuted for work. They had a telegram group chat with two "channels", one for talking, one for bot posts. The telegram bot could be sent a single screenshot of a raid, and it would use OCR to automatically generate an interactive UI for that raid for everyone to see, with all the relevant info, and it would also clear itself up when the raid is no longer relevant. You could press buttons to say you were going, that you MAYBE were going, if you were late, and if you already started/done it, all in single clicks. Tons of options, tons of information, all live updated.
I was bedazzled. That feature singlehandedly removed all attrition from urban social gaming. And it was entirely grassroots. It made me try out making my own telegram bots, and yeah, you basically have the power to make a little app in chat form, even some that feel like CLI commands.
It's been OVER HALF A DECADE and I have yet to see a single other chat application have that degree of freedom where it comes to applications and bots. Some like discord even did whole ass 100% reworks of their bot AP to support the likes of slash commands, and still fall short. And there's none worse than Teams. Teams hates you. Teams spent the prior 2 years before this one basically pointing a gun to our heads telling us they were removing webhooks and pushing back on it whenever they repeatedly get told that's the most insane and dogshit idea ever. And they still did it. There's just no spark in Teams UX. No self-respect. It's a soulless product made entirely as a dumping place of "synergy" with other M$ products. It's reciprocal, I hate it too.
Oh and my local group never go into telegram because they didn't want a new app. It died, but I still kept playing after work without problem. It makes me wonder how fast Teams would die if it wasn't proped up by 365 and Azure subscriptions.
I've been surprised how little support there has been for Teams in the whole AI ecosystem. It seems all developers assume that the whole world is at startups working on Slack when most businesses are on Microsoft 365.
We're finding out quickly that enterprise endpoints are not locked down anywhere near enough, and the stuff that users are creating on the local endpoints is quickly outpacing the rate at which SOC teams can investigate what's going on.
If you're using Claude via Anthropic's SaaS service it's near impossible to collect logs of what actually happened in a user's session. We happen to proxy Claude Code usage through Amazon Bedrock and the Bedrock logs have already proven to be instrumental in figuring out what led a user to having repeated attempts to install software that they wouldn't have otherwise attempted to install - all because they turned their brains off and started accepting every Claude Code prompt to install random stuff.
Sandboxing works to an extent, but it's a really difficult balance to strike between locking it down so much that you neuter the tool and having a reasonable security policy.
I've created an iCloud account for my llm. On my Mac, I created another user account, not an admin, just regular. Linked to the iCloud account. Installed Bluebubble.
And now I can chat with my AI via iMessage, via my Apple watch, or my homepods. It works beautifully.
Also because the aws cli works better, just add an instruction like this to your agents file:
> When performing aws cli commands in terminal always use the `--no-cli-pager` flag to avoid interactive pagination.
Signal specifically is missing any kind of official bot support, cutting off massive audiences from even considering it as an option.
Look, as a software dev myself, I really like that my company lets us use our computers the way we see fit. Pre- or post-AI with no restrictive lockdown. Been there, hated that.
But I totally get the freaking out over "normal devs". The amount of stuff most people think is reasonable, AI or not, is mind boggling. For myself of course I like to just be able to be responsible myself. But as a security team I'd also be freaking out.
Like, the amount of people that find our super boring, totally corporate "security training videos", helpful and insightful and "oh dang I'd never have thought of that!" is mind boggling all by itself. Never mind any actual security training that'd be useful to someone with half a brain. You can literally just click through the 8+ hours of stuff you're supposed to watch / answer / do in 30 minutes.
If so, it can either just shove the full heartbeat file to a smarter model or try to intelligently spread the tasks to the correct models.
I have no evidence that Telegram is an FSB honeypot. What I do know is that if the FSB had a honeypot chat platform, it would look and function exactly the way Telegram does.
After building a proof of concept, we decided that we will only continue Teams integration if anyone is going to pay serious money for it.
This is the first time in a few months I might actually try `claude` cli again to try out this channels scheme.
I ended up just running Claude code in a dtach+ttyd session. Still not the best, as xterm.js has tons of issues with long scrollbacks, but it's at least somewhat _usable_.
It's not an actual MCP. It's just built into the UI.
Anthropic has a lot of stuff that's way more difficult to use because of bad UI, spread over lots of different places.
This kind of thing is so common in groups of people, it's one of my pet peeves. My own family does this in our group messages when trying to make big decisions like who should host thanksgiving or where we should go for a family vacation.
I make it a point to just take charge and tell people that we're doing XYZ now. It usually either results in a decision, or gets the discussion going enough that I can do it again with new information.
I had a look at getting same agent up in WhatsApp but it seems to need FB business acc to even start process, to get FB business you need an FB personal etc. ... looked like too much effort
If you are big into logs, OpenAI might be more your speed. They've got an extremely good logging UI in their platform web app. I use it all the time to figure out what the hell copilot was thinking.
Teams network connectivity is a plain joke. If you use suspend, or frequently change network, the thing will just never reconnect, even though you have VPN alive and all network applications perfectly running.
And the thing is just absurbdly sluggish, only display blurred grey lines instead of text in a meager attempt to look snappy.
Months back I was in a meeting and the dial tone just started sounding like someone was calling me.
I face bugs like this often. It's a pos.
When it is online, I agree with things asides from the "fast" part, actually. But many companies have a secondary service for async comms/chat when being Teams cannot be online, and compared to Slack.
It looks like we found a high executive using company money to buy a product no one wants to use.
It's easy to promote Teams if your secretary is handling it for you and you don't need to suffer yourself.
The other possibility: Microsoft started an astroturfing campaign on HN.
If you hit anything else, feel free to reach out to me personally (email in my hn profile) or via the discord channel we have. Always useful hearing from people who actually tried it and ran want to make it better
So I did spend some time retrying it today, and while I can see some improvements, it is still not in a shape where I trust it enough to replace ttyd.
Some feedbacks:
- The UI is bloated with not-so-useful widgets which take a lot of space. On my S22 Ultra, arguably a screen on the larger side of the spectrum, around 60% of the screen real estate is used for non-chat purpose. If I start to enter text, that jumps to 90% of the screen for widgets, and can see barely a line or two of chat history while typing. Most of these widgets are options that you maybe change once in a while (thinking mode, model selection) but don't need to have all the time above the text input. Same for the "Processing / Stop generation" widget, it takes 20% of the screen.
- The login flow is broken. I have to login two times in a row for it to work. Also if anything goes wrong during login, there is no error message and you're just not logged, and end up trying to login through the shell tab, which is painful.
- I still have issues with answers appearing below the text input, effectively hiding the last lines of the answers. Much less than in the previous version, but it does still happen. Refreshing the page fixes it though, which is a win compared to before.
- Sometimes my last message appears at the very bottom of the chat as if I retyped it. Refreshing the page make it disappear.
- Unclear to me what the right hand side options do. I tried many of them and can't tell a difference.
The comment I replied to said all these great things about Telegram, as if it where a marketing copy, but none of the downsides.
Maybe what sucks here is the experience of running it on Windows. Or maybe it sucks for large meetings? But I never have Teams meetings with > 40 people at this company.
For encrypted group conversation over third-party networks use Signal, or Matrix which try to keep your conversations private.
The fact that Telegram has their support playbooks online, is interesting. Though I would call the following claim a stretch
> As a result, we have disclosed 0 bytes of user data to third parties, including governments, to this day.
Telegram gave more user data to French authorities after founder's arrest https://www.lemonde.fr/en/pixels/article/2025/01/08/telegram...