Excited that we made the #3 spot via just word of mouth, no paid marketing. If you have any questions after trying https://zero.xyz (it's free! tell your agent to install the CLI!) with the $5 welcome gift, ask 'em here!
Just a quick update that as of this writing there are exactly 8600 services indexed and available on the platform now - and growing - and you can browse them all here: https://www.zero.xyz/browse
please update using the following command:
curl -fsSL https://zero.xyz/install.sh | bash
It's awesome! Does what it says.
An ability to add niche tools.
@michael_ludden ahh was thinking more like pointing an agent to zero inside CLI to get it done.
Zero is a no brainer for anyone working with agentic AI. Once you install it, you use your agents as normal, but if it runs into something it normally couldn't do, there's a good chance that Zero can help your agent find one of the around 8000 x402 or MPP listed and stack ranked tools available on the 'agentic web' to solve the problem so your agent can accomplish the tasks right from terminal. Eliminates tons of time and effort finagling config and API key setup for common tasks and lets you get a lot more done right from a single prompt. Zero also isn't charging anything, and you get $5 for free - check it out!
@michael_ludden Congrats on the launch, very cool. How do you deal with tool quality/choice paradox for the agents iro selecting the right tool?
@michael_ludden @zolani_matebese You and the agents choose what tool is best! We help you make an informed decision by scoring each tool, based on agent reviews. Every time your agent uses a tool, it provides a review based on success/failure, value, reliability, accuracy, etc.
@benyamin_chenkirov Well I think everyone wants personal concierge that can do more and more to help them! But with the speed of AI, everything is changing so fast. Exciting times!
@michael_ludden Hey Michael, congrats on the launch. The "4k tools is the headline but selection is the real problem" point upthread is right, and the PageRank-style scoring answer is clever. The follow-up I'd push on is the cold-start and feedback-loop side of that. A PageRank-like system has two known failure modes: a genuinely better new tool with zero agent-reviews stays invisible until someone risks it, and popular tools accumulate usage which accumulates reviews which reinforces rank, so the system can converge on "most-used" rather than "best for this specific task". How does Zero get a brand-new tool its first fair shot, and does scoring stay task-conditional or collapse toward global popularity? For an agent picking among 4k options, "best on average" and "best for what I'm doing right now" can be very different tools. That's the part that decides whether the ranking stays trustworthy as the catalog grows. Following along.
@michael_ludden @arturbrugeman Thanks Artur! Fair nuances you've expressed here - honestly, it's all stuff we're asking ourselves and trying to work through.
"How do new products get a fair shot?"
It's a problem faced by any startup trying to breakthrough, but the agent economy actually makes it possible for new players to beat incumbents. On Google, you'd buy ads or try to optimize SEO, spending tons of money. In the agent economy, it probably comes down to building a better product, clearer docs, simpler procurement for agents, and cheaper prices. Agents don't care about the brand name, they care about results.
For Zero specifically, any new product or brand can work directly with us to optimize selling to agents. We've got experience and can help you win.
"Does scoring stay task-conditional or collapse toward global popularity?"
Task-conditional, with global popularity being considered. Given the agent reviews are per-task, the results shown will be per-task. Zero only shows the relevant tools for the request. And depending on exactly how the request is framed by the agent, Zero may suggest the less popular option (ex. if the agent is specifically prioritizing price, that'll be taking into account).
Lmk if you've got any other suggestions or questions!
@arturbrugeman @daniel_baum Great question Artur. Let us know if you have any follow ups - Dan's response is pretty comprehensive but this is such a bleeding edge tool, we'd love to hear your follow up.
@michael_ludden congrats on the launch this is pretty awesome. How are you making sure the Agents are doing the right tool calls or the services they choose is the right service for the users? how are agents handling tool call failure?
@aditya_rohit great question! Generally you can think of Zero as a tool that agents use to learn about what services are available on the agentic web to help them fulfil a user's request, and to facilitate execusion when the decision to do so is made. The actual decision making about which service(s) to use ultimately lie with the end user and the agent. We try to ensure that Zero provides agents with the best information and context about the services that might solve the problem they're trying to solve on behalf of their users, so that they can make an informed decision on behalf of their users. If a service fails to execute then that is recorded as a negative point in terms of the reliability metadata on that service by zero, and the agent will make a determination based on what the user asked them to do as to whether to try to find a different service or just report failure to the user and wait for input before continuing. Hope that makes sense!
we built this initially for OpenClaw hyper-adopters like ourselves but it works everywhere else too.
creating accounts and provisioning API keys for every service you want your agent to use once is painful. doing it again for every new agent you create is a nightmare to manage (even more so for a company). canceling accounts you didnt use or remembering which agent has which tools is a pain in the a$$
lots more to build here on day zero of the agentic economy (pun intended) but props to the x402 & MPP teams for laying the foundation for us all to build agents with super powers.
The 4k tools claim is the headline but the harder problem is usually tool selection. when an agent has access to everything, it often picks the wrong thing or chains calls inefficiently. Curious how you handle tool disambiguation and whether there's any ranking or context-aware filtering happening under the hood. Also wondering what the auth story looks like for services that require OAuth or API keys across different users.
congrats for the launch anyway!
@fberrez1 Thanks for the question! We help you (and your agents) make informed decisions by scoring each tool, based on agent reviews. Every time your agent uses a tool, it provides a review based on success/failure, value, reliability, accuracy, etc. The scoring then becomes like PageRank.
As usage scales, so will the volume of agent reviews. So the scores become more reliable.
Got any ideas to further improve the system?
Unified tool registry for agents is something we've needed badly. Building RetainSure's AI workflows means stitching together CRM APIs, ticketing systems, and comms tools and each integration is a custom adapter. The 4k tool count suggests a standardized abstraction layer over wildly different auth schemes. How do you handle tools that require multi-step OAuth flows or dynamic credential management per end user?
@anand_thakkar1 it's actually closer to 8k now! No auth schemes involved - purely transactional. Tools that require authentication are not part of what Zero is currently solving for because they'd require additional setup steps. Can you give an example of a service you're thinking about btw?
Whoa, awesome to see it fully live! Will be spending a lot more time with Zero. Congrats on the launch, @michael_ludden and team!
@michael_ludden @alwaysunday Thanks Andy! Let us know any feedback - it's still V1 we're open to anything!
The x402/HTTP 402 bet is the architecturally interesting choice here. Zero is essentially wagering that micropayment-per-call becomes the standard settlement layer for agentic API access, which would be a significant infrastructure shift. The question is about the failure mode: when an agent hits a 402 Payment Required from a service that's not in Zero's index, does it fail gracefully or does it try to handle the payment directly? And is there a spending cap mechanism the developer can set per agent session, not just per individual call?
@binu_george heya! GREAT questions! Gold star haha.
We actually also support MPP in addition to x402, but to answer your question: When an agent hits a 402 Payment Required from a service that's not indexed already by Zero, it is completely up to your agent (or you, depending on how you've instructed it), not Zero, as to whether or not to transact with that service. Zero is a discovery mechanism for such services, as well as facilitation layer for tractions, and generally indexes ones that appear functional, meaning they've successfully run at least one time already and delivered a result that is expected, but if your agent for some reason wants to transact with, say, an x402 service that hasn't yet been indexed by Zero, that would probably put it on Zero's radar to consider indexing. And, to be honest, most services on x402 and PMM are indexed by Zero pretty quickly, unless they're brand new or fail consistently or do something shady.
Micropayment-per-call is definitely already a settlement option for certain API endpoints that people expose via x402 or MPP or other emergent protocols, including ones by Shopify and others, that are compatible with transactional payloads (I pay you I get this thing). Obviously there are some API use cases that can't function in a purely transactional manner, so I wouldn't necessarily say it'll become the only settlement layer for agentic API access - just my opinion here.
The spending cap mechanism currently in place is the instructions you give your agent about how to spend to accomplish a given task, or in general, and the funds you fill their wallet with to use. I personally think of it like an agentic 'allowance' like ya might give a child. ;)
Any follow ups, please ask! Would love your feedback after trying it out.
One operator wrinkle I'd want to understand is the stop boundary. Giving an agent access to 8k tools is powerful, but it also raises the cost of one bad branch if the agent keeps widening the search with no clear receipt. Do you expose enough evidence for a human to see why tool A lost to tool B, what failed, and when the agent should stop instead of trying adjacent tools? Discovery is great, but trust usually comes from the handoff and rollback story.
@keesan12 short answer to the first two thirds of your question "Do you expose enough evidence for a human to see why tool A lost to tool B, what failed, and when the agent should stop instead of trying adjacent tools" is I think a resounding "yes."
Each service is ranked using a ton of metadata from previous runs and agent reviews. This includes how often a request is successful, how long it takes, how expensive it is, how long it has existed, how many runs it has had overall as well as reviews over time, and qualitative feedback from agents and humans that leave those reviews. Agents generally like to leave reviews after runs, and Zero encourages that.
In terms of "when the agent should stop instead of trying adjacent tools" - I think that's squarely in the realm of your and your agent's discussions and configuration. If you tell an agent to spend only a certain amount of time trying to find a solution before moving on, that would, for example, gate when it should stop looking around.
One maybe implicit assumption I want to call out is maybe you're imagining the agent will just be spending tons of money paying for results. I would encourage you to try out the product! That's not how it works, once a service is successfully executed, agents show users the results and then you can decide whether they should re-transacpt using a different/better service. You'd have to explicitly tell your agent to keep trying different services and spending money for it to exhibit this behavior, in my experience. But, again, that's an agent side thing, not a Zero thing.
Great questions, keep em coming! Hope this is helpful - try out zero.xyz and lmk your actual real world experience too! You get $5 to experiment with it anyway, which goes a long way believe it or not.
thanks Deb! Great callout - are you thinking the ability to add tools via the web interface, like a button that gets the process started?