Vishalsai Daswani

Building Agent Manager HQ
Always available, always in budget, always secure Openclaw instances

[Vishalsai Daswani @vsai12]

We’re helping people manage the agent (context, skills, memory) while we handle the plumbing (infra, security, identity).

Your business’ edge comes from your unique personality, not generic AI wrappers and context that everyone else is using.

[Nik Sharma @mrsharma]

Every consumer AI assistant app is basically a closed box doing what you could build yourself in 45 minutes with an agent framework

You'd own the memory, the context, the skills you create from your own usage and actually understand how it works. I'd rather have 1 system with 20 agents than 20 apps with 20 disconnected experiences

Plus, all these apps store your data on their servers with their own rules about your data

[Vishalsai Daswani @vsai12]

@mrsharma 100% agree. More than that, your edge comes from your own context, skills, memory. Relying on a wrapper doesn’t get you in the top 1%.

We’re helping people manage the agent (context, skills, memory) while we handle the plumbing (infra, security, identity).

[Y Combinator @ycombinator]

AI-Native Service Companies
@gustaf

The total spend on services is many times larger than the spend on software, and a lot of those services are already outsourced, which makes them easier to replace with an AI-native product.

We're excited about companies that don't sell a tool to help you do the work: they just do the work.

[Vishalsai Daswani @vsai12]

@vovudebosh Is this real?

I tried using explee and it found all the wrong target customers, and wrote very AI-speak emails that imo don’t build much trust with my users.

[Caspar @caspar_br]

anatomy of a managed deep agent. https://t.co/pnmD06la9i

[Vishalsai Daswani @vsai12]

@gregisenberg Grokbot is cool, but medium-large businesses need more:

1. meet team where they live: email, slack, whatsapp
2. budget-conscious
3. fine-tunable. the competitive edge comes from your control of context
4. multi-user
5. observable/traceable/secure

[Vishalsai Daswani @vsai12]

@anothercohen @Zenomercer @alibrocato Did you try setting up your own opensource harness locally or on VPS?

Instinct seems like a wrapper around openclaw/hermes, without even basic security concerns.

Self-managing will give you much more visibility, control, and privacy.

[Vishalsai Daswani @vsai12]

@xbillwatsonx What does the runbook solve? Is this something that cannot be achieved with improvements to Openclaw itself?

I found it easy to set up in the past, but it's definitely challenging to maintain. And so am curious how this would help.

[Bart Mol @Bart_Mol]

Me explaining AI to friends

[Vishalsai Daswani @vsai12]

Unpopular opinion: @openclaw and @NousResearch could be better than @bot, if it "just worked"

closest opensource alternative to GrokBot: https://t.co/OVonIk34Pz

People will want more granular control over their context and skills, but until then, things need to still work

[Vishalsai Daswani @vsai12]

Finally got my dev env VPS setup using a pre-configured Docker image.

Input: Github PAT
Result: cloned dotfiles, repo, tmux, gh cli, croc, nvm, npm, claude, ready to go

Now I can
1. onboard any engineer in <5 mins
2. go to the gym while my agents work

Am I missing anything?

[Vishalsai Daswani @vsai12]

Plugins are just the first step of what is needed. This is the equivalent of npm packages + CI/CD.

Similar to how we have version control and automated tests across different architectures, we’ll need versioning and evals running across all LLMs.

[GREG ISENBERG @gregisenberg]

Send this to ANYONE on your team using AI agents with Claude/Codex skills:

If I were you, I'd put your BEST AI skills in a GitHub repo, turn that repo into a PLUGIN, and have your team INSTALL it in Claude/Codex with auto update on.

Why?

1. Everyone gets the same AI SOPs instead of 10 versions floating around Slack

2. When one person improves a skill, the whole team gets the better version.

3. New hires can start with your best workflows instead of a blank AI setup (this is a BIG deal).

4. If someone breaks a skill, you can roll it back with version control.

5. Personal skills can stay personal, while team skills become shared company infrastructure (and an asset!).

6. Your best AI processes stay with the company when someone leaves.

7. You can chain skills together for bigger workflows, like titles -> thumbnails -> descriptions -> YouTube publish.

8. You can track which skills are actually being used and delete the ones collecting dust.

9. You get less slop because the agent has real instructions, examples, taste, and process.

10. Your team moves from single-player AI to multiplayer AI.

(thanks to @aiwithremy for coming onto @startupideaspod)

Watch full breakdown here (clearly explained): https://t.co/cvIuFBhOEB

I don't know why I didn't do this before.

The more I think about it, the more obvious it feels: your AI workflows should be version controlled company assets.

It kinda feels like the difference between “we use AI” and “we actually operate with AI".

Enjoy.

[Vishalsai Daswani @vsai12]

@StartUpRabbi @gregisenberg Right now, everything is changing so fast that even @openclaw users can’t be sure their agents are running reliably.

We’re def gonna get there though! 🚀

[Vishalsai Daswani @vsai12]

@StartUpRabbi @gregisenberg In a multiplayer world, we need a lot more guardrails and determinism.

version control on skills, configs is easier with git, but need it also for agent-editable binary files (videos, docx..)

Security (prompt injection), budgeting, backups, and other needs

[Vishalsai Daswani @vsai12]

@threadkillerokc @StartUpRabbi @gregisenberg More than just skills, there’s a lot more infra required for supporting reliable team agents: budgeting, prompt injection protection, uptime, context (skill, plugins, memories) versioning and recoverability

We’re working on that (although only openclaw support atm)

[Vishalsai Daswani @vsai12]

@JoeMcCarley @hrudolph @JoeMcCarley what are you using for archiving your data?
In-built openclaw backups?
Snapshotting the full disk?
Manually copying specific non-version dependent files out?

[Vishalsai Daswani @vsai12]

You should own your agent's context, memory, and skills. That's your edge.

AgentManagerHQ handles the infra: uptime, security, identity, budgeting.

https://t.co/xOd4uppTYn

[Nik Sharma @mrsharma]

I have a lot of respect for anyone building in the consumer AI space…

But why is everyone so excited about these assistant apps that are basically just a black box?

They’re essentially a Hermes agent with skills to be a better assistant … why not just build that for yourself and own the infrastructure?

It only takes 45 min to setup and you’ll have the memory, context, skills for all your agents + the knowledge of how this stuff all works!

Call me old school, but I’d rather have one system with 20 agents vs 20 systems running 20 bifurcated experiences

[Vishalsai Daswani @vsai12]

@mrsharma @mrsharma I'm working on an agent fleet manager that helps users manage the agent, not the infrastructure

Users manage their skills, memory, context. We provide uptime, maintenance, identity provisioning, security, etc.

would love to chat if you're down

[Vishalsai Daswani @vsai12]

@AnomalistG @atomicbot_ai @Aditya_T007 @AnomalistG those tencent prices looking insanely ngl