Click
Managed Agents API that works with any model
223 followers
Managed Agents API that works with any model
223 followers
Managed Agents API you can use with any model. Send a task with an API call, and Click runs the agent and streams back results.
Ask your coding agent about useclick.ai/reference!
This is the 3rd launch from Click. View more
Click
Launching today
Managed Agents API you can use with any model. Send a task with an API call, and Click runs the agent and streams back results.
Click gives you the entire agent stack over an API, with no infrastructure to manage. Sessions are created with everything your agent needs - a sandbox, browser, and web search. Agents run on a harness with the latest innovations, like deferred tool discovery to support 1000s of tools, and Code Mode to use tools with scripts.
Payment Required
Launch Team / Built With
Framer AI AgentsDesign and publish professional sites with AI
Promoted
Hi PH, Aditya here!
Excited to launch the managed agents API that works with any model - BYO models & tools, and run agents over an API. Ask your coding agent about useclick.ai !
Deferred tool discovery is the part I'd want to test. Every StorkPlus store has its own MCP server, so an agent working across several stores could see hundreds of tools with near-identical names like product search and create order. When I create a session, can I scope which MCP servers it gets, so an agent for one store can never call another store's tools?
@ahmed_khaled32 Yes, every agent is provided with only the tools you grant. Specifically, you first create tool connections, which you then attach to an agent. This allows for scoping MCP tools per agent across multiple users. E.g. you can create separate tool connections for the same MCP with each user's credentials and attach it to their agent.
More details in the API reference - https://useclick.ai/reference#tag/tool-connections
Code Mode is the part I'd want to try first — using tools with scripts instead of one call at a time should cut a lot of back-and-forth on multi-step work. The session coming with a sandbox, browser and web search already set up is the other draw, since that plumbing is usually what I end up rebuilding myself. Does Code Mode work with BYO tools too, or only with the built-in ones?
@jcbuilds Code mode works with all tools.
Congrats!!
One of our scheduled agents sat on a permission prompt after a settings change, and we only noticed the next evening. If a Click agent is stuck waiting on something like that, does the API say so, or does the run just look busy?
@julien_zammit There are two kinds of permissions:
You assign a list of callable tools at the time of agent configuration.
To have an agent interactively ask an end user for permission, you provide a custom tool connected to your application. The tool call appears in the API's event stream and session state. Your application decides how to interact with the user.
third launch under the same name is a lot of repositioning in under a year - cursor, then research context, now a general managed agents API. curious what actually carried over on the infra side versus what got rebuilt each time. also "works with any model" is easy to claim until a long session dies halfway through - do you get the full execution trace back when that happens, or just whatever state existed at the point it stopped?
@galdayan Across those products, I ended up building the same agent stack over and over again, before I could get to the differentiating edge of my product. That's what got me interested in the Managed Agents space. A frontier agent runtime that works with any model and can be integrated with an API is what I wished I had.
@adiasg That makes sense, building the same harness three times would push anyone toward productizing it. Still curious about the failure case though: if a long session dies halfway, does the API hand back the full execution trace so I can resume or debug, or just the last known state before it died?
@galdayan The platform handles recovery. The API exposes a durably recorded execution trace so you can inspect what happened.
More information: https://useclick.ai/reference