# Building a Component Create a tool in the component wizard, declare its typed ports, and let Perceivo check the script against them. - Page: https://perceivo.ai/university/orch-05-building-a-component - JSON: https://perceivo.ai/registry/film/orch-05-building-a-component.json - Kind: film - Revision: 3 (published 2026-10-18, produced 2026-10-15) - Previous revision 1 (published 2026-10-07): https://perceivo.ai/registry/film/orch-05-building-a-component/r1.json - Previous revision 2 (published 2026-10-11): https://perceivo.ai/registry/film/orch-05-building-a-component/r2.json - SDK version: none ## Validation - SDK: not applicable (UI walkthrough; no component or orchestration is offered for download) - Platform: not applicable (UI walkthrough; no component or orchestration is offered for download) ## Transcript [0:00] Introduction to Orchestrations, Episode 5: Building a Component [0:06] Welcome back to our Introduction to Orchestrations series. [0:09] Last time, we covered the component types, and the rules for connecting them. [0:15] Now we'll build one: a small tool, with a typed contract that Perceivo checks before it can ever break a run. [0:23] Every component in your organization lives here. [0:26] You can import one from a file, or create one from scratch. [0:30] Let's create one. [0:32] The wizard has five steps: general details, ports, the script, who can use it, and validation. [0:40] We'll give our tool a name, and a description our team will understand at a glance. [0:45] We covered the component types last time. [0:48] We're creating a tool. [0:50] And we'll accept the default appearance. [0:53] Ports are contracts. [0:54] They guarantee what inputs a component accepts, and what outputs it provides. [0:59] They're what makes components safe to snap together. [1:04] A tool takes exactly one input port. [1:06] We'll call it input. [1:08] The request specifies what the tool expects to receive from its caller. [1:12] The sender has to pass what your script requires, or your script will likely fail. [1:17] That's why you declare the port in your manifest, to announce to the world what you will accept, and what you will return as a response. [1:25] Perceivo validates those ports at the time the connection is made. [1:29] But it also validates that your script implements what the manifest says, when you create the component. [1:35] To illustrate how that works, we added a string field called text, to the request. [1:41] And for the response, we promise to return an integer called word count. [1:46] By the way, if you'd rather not use the form, all of this can be defined in a manifest. [1:52] In this view, your ports are in a YAML format. [1:56] If you develop locally, or use a coding agent, this is the form you are more likely to see these definitions in. [2:03] But some people like the graphic interface, so we provide both options. [2:09] With the interface defined, it's time for the business logic. [2:13] The wizard can generate a starting script for the type. [2:16] We'll replace the scaffold with our business logic. [2:20] Notice the key parts. [2:22] This wrapper means it extends the Tool class. [2:25] It has a run function. [2:27] That is required. [2:28] It reads text, the field our port declared, and it returns response data. [2:35] Every asset you create in Perceivo has its own permissions. [2:39] Roles is where you assign them, to groups, users, and service accounts. [2:45] Validation runs automatically when you reach the last step. [2:49] It reads the script, and checks it against the manifest. [2:53] And it caught something. [2:55] The script returns a field called result, which our port never declared. [3:00] And it never returns the word count the port promised. [3:04] If the script and the contract disagree, you find out here, not halfway through a run. [3:10] So we go back and fix it. [3:11] We read the text from the request, count the words, and return the count. [3:17] Notice the names: text and word count, exactly as the port declared them. [3:22] Now the script and the contract agree. [3:25] Validation passes, and we create the component. [3:29] And here it is, listed with our other assets in the palette. [3:33] Wherever you wire it in, the canvas checks that connection against this same contract. [3:38] Everything the wizard just did, your coding agent can do with the SDK. [3:43] It scaffolds the component, validates it, runs it locally, and exports a package you can import with the Perceivo API. [3:52] And this structure works for your agent. [3:55] No boilerplate cluttering its context. [3:57] Less to hold in its head. [3:59] It spends its attention on the problem, not the plumbing, and it performs more reliably. [4:06] So: define the contract, write the script, and let validation prove they agree. [4:12] That's a component you can rely on. [4:15] Everything in this series is open source, with example components and every example that follows in the Perceivo SDK repository. [4:25] Next time, we reuse work: whole orchestrations inside other orchestrations, and templates. [4:32] See you there.