I was a little surprised to see a Telegram integration rather than Slack or Teams, given Anthropic's enterprise-first posture. But then I looked it up, and it turns out Telegram dwarfs both, at around 1bn MAUs, vs 50m and 300m respectively! I had no idea - reminds me of the time I found out Snapchat has 2x the userbase of Twitter.

Claude is leaning into the idea of a local "session" being the host where everything connects.

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.

This is actually great for *claws. When Anthropic changed their T&Cs to disallow using claude code oauth tokens in the Anthropic Agent SDK, you had a choice between violate the terms or pay a lot more for the model inference using an API key from platform.claude.com instead of claude.ai.

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 am not sure how I feel about all these hype-driven tools honestly, especially considering they are super janky since probably rushed out with Claude Code.

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.

What these 'channels' do is essentially why I was running a nanoclaw at work: triggering a claude code based on events and getting feedback/review/analysis which nicely closes the loop with other agents.

Not sure why it has to be an mcp, but will be trying this out asap.

At this point the limitation is even requiring a terminal in the first place.

Claude Code daemon mode in background when?

Claude caught up pretty quickly. I think OpenClaw’s core value is the channel, heartbeat, and the open-source ecosystem.

I just created agent-http that leverages the channels feature to enable you to wrap claude code with a http api. This provides an identical API to Agent API (https://github.com/coder/agentapi) that relies on terminal scraping to achieve this. Now you can interact with claude code in a headless manner using your subscription. Previously I think you had to do this via the Agents SDK which relies on api token use.

at this point anthropic is dogfooding us a new product every week just to see what might stick - doubt a lot of the features/products they've rolled out will actually be around or supported in a year

Very cool!

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 enabled the github connector in claude web interface.

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.

This feels like a response to openclaw (and openai's hiring of the lead).

I don't understand how this can be economically viable. If this takes off, it will allow businesses to use openclaw-like functionality at non-api prices (pro, max).

Nice, i've been waiting for this capability to show up. I've added support to my side project llmsg.com, here's a video of it in action https://x.com/sho/status/2034898928618152412

This is one of those features that sounds small but actually changes how you structure things.

Not having to restart or rebuild context every time makes a big difference once systems get more stateful.

It would have surely taken less time to just set up notifications for the Claude Code app? Are they ever going to do this? It's baffling to me that they're just skipping over letting you know when a task is completed... this is basic stuff.

This is fantastic. There are a ton of use cases where you'd want to be able to build an integration that hooks back to your running agent session. OpenClaw has this today, but it's pretty janky. Hopefully this is coming to Claude Cowork as well.

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.

I was making a telegram to Claude via tmux capture-pane and send-keys, this will be so much nicer. Also sounds like something that addresses some of what Steve Yegge said was missing for agent to agent communication as well.

I like it but it's one conversation at a time, no ability to manually compact the chat... I guess it's an early version still.

I was reading through CC MCPs docs ... MCP Notifications kind of did this right? The server/client could update each other automatically. So there was already this channel-like communication.

This is like that but instead of the server/client sending messages it's you.

I’m struggling with how useful this is in an enterprise setting where a security team will be at least slightly annoyed at devs hooking into their personal machines via untrusted chat… but OK, I guess. Myself, I have pretty dropped WhatsApp support in https://github.com/rcarmo/piclaw and have made sure its auditable through a proxy…

This is an interesting direction.

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.

But can it edit tabs? [1]

[1] https://github.com/anthropics/claude-code/issues/11447

The convenient thing about using Claude via Telegram is that you can provide all of your private and proprietary information to US intelligence and Russian intelligence at the same time. (Telegram is not end to end encrypted.)

I've been looking to build something similar to this so this is very timely!

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.

Claude getting clawed.

Interesting to see it took them so long to implement this. Claude was super limiting without the ability to have a scheduler or a connection to events

Will need to push the entire session as an object event.

I could not answer yes/no over Telegram so this makes no sense. Why is this gigantic company shipping these vibe coded changes to quickly and partially?

Channels are a natural fit for real-time data pipelines. I've been building MCP servers that push structured data from multiple sources (news feeds, social media, academic papers) into AI agent sessions.

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.

It’s going to be fascinating to see what kinds of malicious execution and exfiltration this enables.

This is the first time I've seen MCP's push capabilities come in handy. I'm not much of an MCP nerd though so I don't know much. But when I read the spec it looked extremely over engineered partly because of the 2 way nature of it.

This looks super super useful.. I'm making an agent to agent chat tool (that I think is actually ready for testing, so please check it out) -- https://chat.corpo.llc/ or https://github.com/corpo/qntm -- and the difficulty of getting claude to check and respond to messages is real.

Basically the Claude CLI is the operating system is the product vibe I get right now.

This is exactly what I planned to figure out how to do: maintain an instance of Claude that can accept triggers that become tasks.

It’s quite basic if I am using it correctly! It expects certain commands to be still approved on main machine.

been running something similar with openclaw for a while now - github webhooks triggering code review, slack messages kicking off tasks, etc. nice to see anthropic building this natively into claude code. the telegram/discord support is a smart call too, way more devs hang out there than people realize.

OpenClaw approach has moved into frontier companies I see -

Lol made the same thing using claude earlier: https://www.viahuman.xyz/ They are gonna implement everything, aren't they?

Really surprised for the frequent innovation of Anthropic

Just as I started to move away from events.

Does anyone else have issues opening Claude.com domains on iOS? It’s infuriating I can never open documentation or the usage page or account management portal on iOS on Safari. Works fine on a laptop. Mac, Windows, or Linux.

Honestly not sure what I would be using this for when there's Claude remote control? Is it because you can script the telegram bot to send messages at regular intervals? But Claude has a /loop as well, so I'm still confused.

Claw-ification

we have OpenClaw at home

(and it may be better)

Tailscale + tmux + Moshi. You're welcome.

finally! I'm building an app that's essentially a "sidecar" to an llm subscription and works via mcp and has a web ui to make reviewing deliverables easier, uses the user's subscription for intelligence instead of requiring to pay for tokens inside the app. The problem until now is I couldn't trigger AI work from the web ui, that limitation will be soon gone, it fixes a huge ux issue for me, I honestly thought it would happen sooner but I'm glad the industry is catching up.

so its a webhook

Imagine if they were able to support iMessage.

[dead]

undefined

[dead]

[dead]

[dead]

I also get the impression this is way more complicated than it needs to be. Or maybe it's simple and they keep inventing new terminology for stuff that basically already exists. The crypto bros did the same shit. Like, bidirectional communication has been a thing for decades. We're just changing what we call the client and the server? And the protocol is just strings the bot on the other end is a little better at reading?