Activity
Mon
Wed
Fri
Sun
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
What is this?
Less
More

Memberships

AI Agents | OpenClaw

101 members β€’ Free

4 contributions to AI Agents | OpenClaw
Agent (and other parameter) Editing
Now that I've been tinkering with OC for a few weeks, and I (or my agent) have broken more things than I care to admit. How does everyone tend to makes changes to their agent(s)? Even the CTO Factory Agent, while much better then my original agent, still breaks stuff from time to time. I been mostly using the agents themselves to make config changes, but unless you're on a top tier model (and even then) it can be problematic. I also use the OC terminal for some things, but that tends to break some of the customizations I'm running. Then I also sometimes edit the .json files directly. Mainly to recover from a catastrophic failure (i.e. OC won't start). I haven't used the WebUI for editing, but that IS how I mainly interact with my agents. But aside from the chat, which isn't great, the rest if the WebUI is pretty clunky IMHO. So just looking to see how other people do it.
Poll
3 members have voted
1 like β€’ 14d
@Vlad Praskov Since I'm just really playing around with OC to see what I could do with it, before I ended up spend a ton on hardware and API credits, I've been running it in a locked down container on my home server. So that presents a few different issues with trying to follow most guides. I've managed to get most things working, and one of CTO Agent's first tasks was to build a "coding" agent to replace the Claude Code/Codex requirement, which actually worked fairly well. I then put it to work helping me "fix" my other agents, which has also worked fairly well, but I think because I haven't figured out how to enable sandboxing in my setup without it bricking everything, the agents can be a little fast and loose with some of the core files (i.e. openclaw.json). I was actually quite amazed that CTO got the voice output working by just pasting in the Part 8 article that just went up. Because for Part 7 we had to do a few workarounds, like installing whisper and ffmpeg locally inside the container.
Telegram vs. Discord
It seems that most guides, not just on here, seem to prefer Telegram as the default interface to interact with your agent. I frankly hate Telegram. The few times I've used it, I was just inundated with spam and scammers, and I'm much more comfortable with Discord. Granted I've never run my own server on either until now, so maybe that would make Telegram bearable. That being said, I've found myself working primarily directly with my agent(s) via the WebUI chat or doing things myself via the terminal (either my server or the openclaw container). Now that I seemed to have tamed my main agent's persona thanks to help from some of the guides on here, I'm ready to start trying to implement some of the more advanced concepts (CTO Agent, Notes Agent, etc.), and I'm wondering if I should pivot to Telegram or just stick it out with Discord. I've already had my agent (Skippy) download and review the code of the CTO Agent, and it says migrating it to Discord "shouldn't be too difficult." So before I start banging my head off that tomorrow, I thought I would see what other people think.
0 likes β€’ 18d
@Dima Citizen Not yet. I downloaded the source, but I don't think I can just run the script and let it install due to my container's setup and lockdown.
1 like β€’ 16d
@Dima Citizen Finally got it shoehorned into my deployment, and so far so good. Going to have to see what I can do without a coding agent, but it already started trying to figure out a workaround.
Quick one for everyone optimizing OpenClaw spend:
Step 3 is live πŸ”₯ : Option 1: Aggregators as your token control panel This is the practical setup for OpenRouter: how to find strong :free options and lock your future model IDs for config. Drop your questions below and I’ll answer each one. 😊
2 likes β€’ 21d
One thing these posts haven't covered yet is how to actually get those models into your config, which isn't that hard, but the way I went about it via the openclaw terminal ended up replacing my primary model with "openrouter/auto" After a few prompts I realized was NOT a very good default, as each prompt by design ends up going through a different model, and now your agent has split personality disorder. So while having some :free models as fall backs is probably a good idea, make sure you don't leave your primary on "openrouter/auto" I'm also waiting for Part 5.3 on setting up a Gemini account, though I have read that Google may not like it being used for an agent.
Idea β†’ deployed β†’ running capital
All from a Telegram message. On my way home from the ATO conference in Miami last week. I built a trading bot I’ve been thinking about for years. No dev team. No weeks of work. No back and forth. I opened our CTO agent gave it instructions and deployed everything through Telegram. The bot is now live executing trades tracking performance and managing itself. Too early to judge the strategy but that’s not the point. The real shift is this: Any idea you have can now be turned into a working system in minutes. Not someday. Right now. I built it just by promoting the CTO agent.
Idea β†’ deployed β†’ running capital
1 like β€’ 24d
Trading onchain or via a CEX's API?
1 like β€’ 22d
@Dima Citizen Good so far if you seeded it with "$3000", but fees are killing it's profits. Would be interesting to know what it could do tied into Hyperliquid or Lighter. πŸ€”πŸ€”
1-4 of 4
Doug Keim
2
3points to level up
@doug-keim-1625
A retired engineer learning that he can code again.

Active 13d ago
Joined Mar 25, 2026
Adelaide, South Australia
Powered by