Openclaw - your free open-source AI personal assistant (formerly Clawdbot & Moltbot)

[robi]

Ooh this would be great as a custom App.
Have it be a specific part of the automated agents tiny company.

[marcusquinn]

@robi ideal to be run in a container, too. Keep it out of too much trouble!

[robi]

@marcusquinn right, people have been giving this full machines, Mac minis, MBPs, etc.. yet this is perfect for containers and paired with Ollama for local usage too.

Folks on Twitter are exploring all sorts of configurations. Fun to just watch the progress..

Keep up here with the Clawdbot Report:
https://x.com/Scobleizer/status/2015346738517942628

Link to Google Doc:
https://t.co/OVwapPLxfq

[robi]

I attempted an install in the LAMP app, and it works after a memory adjustment, however my browser terminal keeps resetting before I can complete setup and further testing. Flaky WiFi too ;-/

A few helpful notes if you are attempting this:

After installing the LAMP App, bump the memory past 500MB, otherwise the installer will crash, running out of the default memory setting. 1-2GB should suffice.

Make sure you're in /app/data and set it as the HOME.

cd /app/data
export HOME=/app/data

Then just run the installer script from the website:

curl -fsSL https://clawd.bot/install.sh | bash

UPDATE 1:
If installing brew, the install.sh script will have to be modified to use /app/data instead of the hardcoded /home/linuxbrew path.

[timconsidine]

Clawdbot seems viable (and interesting) for Cloudron so I have started packaging ....

[micmc]

@timconsidine said in Clawdbot - your free open-source AI personal assistant:

Clawdbot seems viable (and interesting) for Cloudron so I have started packaging ....

WHAT? ...

[marcusquinn]

@timconsidine Significant note on the security advice on this one:

• https://docs.clawd.bot/gateway/security

Needs to be very much locked-down by default, given the capabilities.

[timconsidine]

@marcusquinn yes, critical - looking into 'surfacing' all of that, and exploring whether it is possible to invoke sandboxing out of the box.
But YouTube is overflowing with videos on security issues, may put some in the POSTINSTALL

[marcusquinn]

@timconsidine Yup. At least link to that page in the readme/instructions.

[zonzonzon]

Thank you so much, this bot is really great. Seamless integration with the existing messaging system offers many use cases.

[marcusquinn]

@zonzonzon I can see uses for many of these talking to each other and delegating tasks.

[timconsidine]

to those wanting the package :

• please do your research now while waiting about security issues, and sandboxing. There is tons of online blog posts and YouTube videos. Just so you can be clear.
• getting close to a package, but not quite there yet

[robi]

Wow things are moving so fast..

Anthropic complained Clawd is too close to Claude, so they rebranded to Moltbot. The next evolution, ha brilliant.

https://github.com/moltbot/moltbot

[timconsidine]

Hoping to release Moltbot (was Clawdbot) for Cloudron later today.

But off to hospital briefly (hopefully) so bear with me.

[jdaviescoates]

Hope your hospital visit went smoothly @timconsidine

[timconsidine]

Yay !
Later than hoped, but done now.

I have packaged Moltbot for Cloudron (was Clawdbot) including its browser GUI.
Repo : git.cloudron.io/timconsidine/cloudron-moltbot
Built public docker image : tcmbp132021/cloudron-moltbot:latest
Available on ccai.appx.uk (or your own CCAI-P but re-import the master catalogue)

Frankly the GUI is not fully developed, and still requires manual configuration in the (ugly but functional) CLI.
But hopefully with the fast pace of development, it will become 100%.

I'm putting this out with an ALPHA status, because I am still testing it out, so don't really know what works and what doesn't. Equally I know there is a lot of interest in it, so rushed it out as ALPHA.
Maybe the 'bleeding edge' / 'edge of seats' brigade can help me out

TBH AgentZero (https://forum.cloudron.io/post/119353) holds more interest for me, so not currently sure if I will use Moltbot in practice, but as it's been getting lots of profile, it's fun to explore and compare and discover its capabilities.

Many of you wanting this on Cloudron are probably 2 steps ahead of me in terms of understanding and configuring Moltbot for real use.
So feedback and your guidance is welcome.

---

Side-note on Packaging : I lost time packaging this because I played with Claude Code (CLI and VS-Code plugin), then Opencode (also CLI and VS-Code).
TBH there was learning time on those, and neither cut it fully for me.
Opencode is good, but didn't perform as well as my standard dev setup.
That's probably an issue of models used, but none performed as well as well as my TRAE dev assistant on "auto" model selection.
I will remove Claude Code from my setup as the only attraction was using it with Ollama models, but I can do that with Opencode or even TRAE.

[robi]

The latest evolution to stay is OpenClaw as it retains its roots and is copyright free.

More here + backstory:
https://openclaw.ai/blog/introducing-openclaw

The gateway really is the cool interface on almost any chat app.

[timconsidine]

@robi openclaw is much better name
I just hope the code is much better and finished

E.g.
“We support iMessage”

• “cool”

“Ah, but it’s a WIP”

• “aarrgghh”

I’m excited about Moltbot/OpenClaw but I reckon it needs another couple months to be complete / stable

[robi]

@timconsidine I hear you. The experience is so different if you choose an option that isn't as polished yet.

I would suggest Telegram or Discord to start as those work well. Haven't tried others.

[timconsidine]

@robi I have Telegram working. Need to configure some / tasks and skills to make it productive.