Launched this week
AgentKey
One-stop live data marketplace for your agent
1.1K followers
One-stop live data marketplace for your agent
1.1K followers
AgentKey is a plugin that connects your agent to live external data in one command. Install it into Claude Code, Codex, OpenClaw, or any MCP-based agent and instantly unlock access to search, web pages, social platforms, finance, e-commerce, business and crypto data. No integrations. No setup. Auto failover keeps workflows running.
Ada.im
Tried it yesterday with codex. From install to first live twitter search was maybe two minutes. That's rare for anything involving third-party APIs.
AgentKey
AgentKey
@s_cen With an LLM in the loop, two minutes means the network part barely showed up. It's a standard remote MCP server under the hood, point and go, and it answers as a plain API too if codex ever isn't your ride.
Is there a spend cap I can set per key? Slightly nervous about an agent getting stuck in a loop and burning credits overnight.
AgentKey
@crystalmei Not yet, honestly. And it's a good enough idea that it's going straight on the list. An agent looping through credits overnight is exactly the failure mode we should have a guardrail for.
AgentKey
The tool discovery part is really interesting. As agents get access to more tools, managing context efficiently will become a huge challenge. How does AgentKey handle custom/private APIs?
AgentKey
@nithikanb Sharp framing on the context problem, that's exactly why the discovery layer exists. Private APIs aren't in the catalog today, it's public curated providers only. Most people just run us next to their own MCP servers in the same agent and split the work that way
AgentKey
@nithikanb https://docs.agentkey.app/concepts/discover-describe-execute This might be helpful to you! We’ve provided a detailed overview of some of the technical implementations behind AgentKey, including tool discovery, context loading, and execution. But Private APIs are an interesting topic; we haven’t covered that part yet. At this stage, the marketplace mainly features providers and vendors for read-only scenarios.
"one account, one bill" is the actual pitch here, not the data breadth. when auto-failover kicks in mid-task and routes to a different provider, does the agent get any signal that the source changed, or is it invisible and just hope the schema matches?
AgentKey
@sabber_ahamed Neither, actually. A mid task switch starts with a visible failure: the agent gets it along with a set of same capability providers, chooses one, and pulls the new tool's schema before the retry. The swap is in the agent's hands, schema included. Between calls our health checks quietly keep bad providers out of rotation, but in-flight responses never change sources.
@lxcong schema included is the detail I wanted, glad it's not hidden. does the agent's context get a note that this is a fallback provider, or does the swapped call just look identical to a normal one downstream?
AgentKey
@sabber_ahamed Self documenting at the agent layer, the failure and the candidate pick are already in its context, and the retry goes to a differently named tool, so nothing about it is anonymous. The call itself is mechanically normal though. For consumers downstream of the agent who never see the transcript, today the trail lives in the console log, and the explicit per-call fallback marker is the thing we already put on the list upthread.
AgentKey
@sabber_ahamed Great topic! We're actually deep-diving into "smart routing" internally right now, including better health checks, fallback mechanisms, and schema returns. AgentKey is just getting started, but we believe these features will be significantly enhanced very soon.
With 1,800 endpoints and growing, how do you decide which providers to actually add versus turn down?
Congrats on the launch!
AgentKey
@jared_salois Thank you Jared! We handpicked our first batch of providers based on the most common use cases—like search, scraping, socials, finance and more. Down the road, we plan to open up a fully marketplace-driven onboarding model. We just need a bit more time to bulletproof our infra and design the vetting process, so stay tuned! Our ultimate goal is to turn this marketplace into a thriving, bustling ecosystem.
AgentKey
@jared_salois pull, not push, for now, we add what users actually ask for, vet it, and skip whatever would just pad the number. The longer arc is a proper two sided marketplace, providers submitting capabilities and us running the rails and the vetting rather than hand picking. Today's curation is the bootstrap, not the end state.
Really interesting approach. One question: if I already have my own API keys for some providers, can I use those with AgentKey, or does everything go through the AgentKey account?
AgentKey
@nayan_joshi Honestly, we hadn't thought about that yet, but that's a brilliant point! We might actually support this down the road. maybe letting you plug in your own keys in the dashboard, or even just telling the Agent via natural language. That way, you wouldn't be billed for using your own keys. Appreciate the feedback!
AgentKey
@nayan_joshi All through the agentkey account today, and honestly that may stay the answer. The value is in the pool being managed end to end: health stats, QoS shaping and failover only work on keys we operate. Your existing keys don't go to waste though. They can sit right next to AgentKey in the same agent, direct connections for the providers you've already invested in, us for everything else.
Super cool! Is it flexible enough to bring in data from specific sources that I already have access to for eg. as a student I used to have access to pitchbook but couldn't use it in the context of an agent because of integration limitations even if I have a subscription.
AgentKey
AgentKey
@margharitha Not today. The frustration is legit though, you pay for access and the subscription assumes a human reader. Public pages the catalog's scraping tools can cover, but a login gated licensed database wants an official integration rather than a workaround. That's where the marketplace direction points: sources like that onboarded through their own channels, agent ready.