Office Hours · 5 November 2026

OpenLint Office Hours #6

Summary

The sixth OpenLint office hours confirmed the project is ready for pilots, and reset the roadmap’s priorities: compatibility and trust before new features.

Status. openlint 1.1.3 is on npm and behaves like Spectral. Miguel Quintero finished the TypeScript 7 work and starts the MCP server (#13) this weekend, with a demo planned for next week. The Learn section now has its own site, learn.openlint.org, for deeper guides on rulesets, extends and aliases, while the docs stay minimal. rules.openlint.org is a generated catalog with one page per official rule. The blog has posts on migrating, the fiscal host, aliases and the Dutch government’s move, plus an RSS feed. A second donation came in, from Steve Willmott.

Rules architecture. Kin presented openlint#40: how formats, rulesets and standards fit together, including MCP, A2A, agent skills, CAMARA, TM Forum and FHIR. He raised a concern: most teams use the default ruleset unchanged, so modularity shouldn’t confuse them. One idea was a “batteries included” default with extra rules ready to switch on.

Compatibility first. Dimitri van Hees explained that the Spectral ruleset is written into the Dutch API Design Rules standard. Every Dutch government organization that used Spectral now uses OpenLint, but many people use vacuum. Because vacuum isn’t fully compatible, Go users copy the standard’s ruleset. He asked to prioritize a next version of the spec that works with both Spectral and vacuum, even as a draft that people can test in pipelines. He’ll write up the differences as a matrix. The main one: Spectral relies on JavaScript’s JSONPath Plus and regular expressions, where vacuum follows the standards. x- properties, such as a rule category, are a way to add features without breaking compatibility. Kin will build a conformance suite with a visual dashboard of how OpenLint, Spectral, vacuum and other tools line up, before anything new is added. He’ll also reach out to SmartBear, Redocly and other tooling vendors.

Performance. Nangiale Shir of Open Resource Discovery (SAP), invited by Pavel Kornev, uses Spectral at the core of a validator that runs large pipelines. What matters most to them is performance, and how quickly the community reviews their opt-in caching PR (#39). Kin will add a performance and housekeeping section to the roadmap and review the PR.

Funding and governance. The Dutch standardization board wants to fund maintainers in next year’s budget, possibly by freeing up dedicated people rather than paying Open Source Europe directly. Kin is writing grant proposals for the conformance suite, the spec, Learn and rulesets. OpenLint is ready for pilots and design partners now, with the official launch at API Days Paris in December.

Agenda and discussion on GitHub · Watch on YouTube

Transcript

00:00:00Before the start: Small talk while people join. Andrzej is in Barcelona; Dimitri joins from a train, coming back from a conference; the US has changed its clocks.

00:01:37Kin Lane: Thank you first off for the post yesterday, Dimitri. Very good, and very appreciated.

00:02:09Dimitri van Hees: Good to hear. I just came back from a conference. I didn't talk about OpenLint, but I still have my badge on. Joost is giving another talk right now, so he can't be here.

00:02:34Kin Lane: Thanks for spreading the word, and thanks for being our cornerstone supporter. Lots of good things are going to come out of your support. Let's give it another minute for everyone to join.

00:03:38Kin Lane: Let's get started. I'll share the agenda. Starting with the recap of last week: we're up to 1.1.3 on npm. We're not doing anything new; we're tightening down, cleaning up and organizing. The goal is that everything works just like Spectral, and all my tests are still good for my tools. But I think we need to start investing in a test harness to keep ourselves honest. TypeScript 7: Miguel, anything to share?

00:04:38Miguel Quintero: Not much. Following Jakub's and Phil's instructions, I wrapped up the big TypeScript work, so we're at a level where we can start adding things. My next focus is the MCP server. I saw your feedback, and I'll start this weekend, put something together, and maybe give a quick demo next week. If I see a few bugs here and there, I might pick something up as well.

00:05:30Kin Lane: I found a small bug in some references from a global copy-and-replace that didn't do what it needed to; Phil and I fixed that. I split the Learn section of the website out into its own repo and site. The difference between Learn and docs: docs are the minimal, what-you-need-to-get-up-and-running, and should stay conservative and refined. Learn is the deeper dives, where we can constantly iterate; it'll be the education center. I added two or three new sections as I worked through rules, modularity and extending rules, because it's been a year or so since I worked with rules at scale in an enterprise. If there are other sections you'd like to add or see added, submit issues on that repo.

00:07:26Speaker [?]: So this is from Dimitri's point last week: someone getting started goes to the docs, and to move beyond the standard ruleset, they go to Learn, click on rulesets, and pick up the community-based rules. This looks pretty good. I'm sure there are rules split by vertical or company, and this would be the place where someone can copy, paste and experiment.

00:08:08Kin Lane: We'll talk a bit more about how that learning has moved some of this work along. I also added a separate repo called rules. Right now it's a single filterable list of the rules in the core rulesets in the tool, OpenAPI, Arazzo and AsyncAPI, plus any rules in the spec document. rules.openlint.org is generated, so don't edit it; it pulls from the tool and the spec. It's meant to be one generated catalog of all the official rules used across the tool and the spec.

00:09:03Speaker [?]: It's a really nice reference. I think that's really useful.

00:09:09Kin Lane: Thank you. I want a single canonical URL for each rule. I'm still working out how rulesets will fit into it, but all the core rules have a home. I added blog posts on migrating and on Open Source Europe being our fiscal host. We got another donation, $25 from Steve Willmott, formerly of 3scale, so I'll keep knocking on doors for money: my original thousand euros plus 25 more. I did a story on aliases, drawn from the aliases work in Learn; I'd never used aliases, so they're brand new to me. I did a post on Dimitri's blog post and the Dutch government's support, echoing what they said. I added an RSS feed to the blog, licensing, privacy and terms, and fixed some broken links. Last week's transcript is in the meetings section of the website.

00:11:13Kin Lane: I wanted to get my head around all this to prime the next work, the roadmap around rules. Right now we have the engine, the default rulesets, and now this rules catalog. But how do we organize the different formats? How do we manage the different rulesets, and how do standards work? How do we add new formats? MCP is top of my list, then A2A cards. Then harder ones: agent skills and Markdown will require changes. And how do we separate things when we have CAMARA, TM Forum and FHIR, which have OpenAPI as well as AsyncAPI? We also started talking about experiences, like documentation and mocking: how do you get modular about extending based on your goals, and help people be more precise in what they're trying to do?

00:13:02Kin Lane: I have concerns here, though. From what I've seen, maybe 10% of people extend, learn and build their own rulesets. Everybody else takes the default out of the box. Maybe the default has a bunch of commented-out things people can light up. We need to think about how we do this, because we don't want to confuse people.

00:13:34Speaker [?]: Kind of like a batteries-included story.

00:13:39Kin Lane: Yeah, exactly. Then distribution: how does Git work? I liked what was said last week about OpenAPI managing its own rules. I could see an OpenLint rules SIG in the OpenAPI Initiative eventually, and I'd like to work with CAMARA, TM Forum and others to help maintain theirs. Ideally those live in our org, but we have to think about it. I'm going to start crafting some of these, which is why I need naming standards in place, so I can start organizing the repos. I'm very hands-on; I have to see it before I propose it. That's issue #40, still a work in progress, rolled up from the Learn and blog work.

00:15:00Kin Lane: Now that the table is set, with the spec, the toolset working, and the base rules and rulesets, what does our roadmap look like? This is where things get harder as a group, and for governance: what do we prioritize? Why would you change from Spectral to OpenLint? We have to pick things that are really relevant to people. I think modularity and reuse, new formats, industry partnerships, the rules registry with good discovery, and a modular Git experience speak to that. And I really want to build on the Dutch government's existing style guides, which we're working through now, with them as an overall cornerstone partner.

00:16:32Dimitri van Hees: From our point of view, a couple of things. Basically every government organization that used Spectral is now using OpenLint. For them, the reason to switch is already that it's maintained; that's reasonable enough. However, a lot of people still use vacuum. This morning before the conference I talked with a member of the standardization board. The Spectral ruleset is now part of the formal API Design Rules standard. I didn't choose it, but it's written in concrete. So we have a dependency from a formal standard to a Spectral ruleset, and that's something we don't want again. And because the ruleset isn't fully compatible with vacuum, people who use Go and want vacuum create their own copy of the formal standard's ruleset, which we don't want. We want the ruleset as part of the standard, so we need to address the differences between vacuum and Spectral.

00:18:19Dimitri van Hees: So I'd really like to start making the spec compatible with both vacuum and Spectral. Even if it's not finished yet, we can get the people now using vacuum on board. They'll say "we miss these features", and we can introduce them in a draft of the next OpenLint spec. That would make onboarding people in the Netherlands much easier. And the fact that the government is involved in this community, through Joost and me, means we're able to fix things if we need to, and it gives users a lot of trust. Even though we're a small country, there are many organizations at all levels, and they all rely on these rulesets. So when it comes to prioritizing, working on the next version of the spec, perhaps even on a branch, so they can start testing it in pipelines, would be a great help for us.

00:20:32Kin Lane: I heard a couple of really important things. One, we can't just abandon Spectral as a toolset or as a ruleset, because Spectral is baked into things. We need to be thoughtful about how we include Spectral tooling in what we do, and stay aware of what the CLI is up to, as well as wider Stoplight and SmartBear tooling, and similarly vacuum and what Dave is building. I know you're going to work on that, Dimitri. I also need to reach out to Redocly and other tooling vendors and come up with at least a communication structure: their understanding of what's happening, their level of interest in being involved, and their awareness that we'll stay compatible with their tools. And we need our test harness in place, with a very visual way of showing that every release stays aligned with these other tools and these forks of the rules. When it doesn't, we're honest about it and say this is a new path we're taking. We need to be very slow, thoughtful and communicative about those things.

00:22:24Speaker [?]: It almost sounds like there should be a section on compatibility, on the docs site or on Learn. Someone hears about OpenLint, goes to the docs, goes to Learn, finds a rule, but they use vacuum. Is it compatible? Do they copy and paste it and then mess with it?

00:22:53Kin Lane: I'd say this belongs in the spec or in Learn. The tool and its docs are less concerned with it; the tool does what it does. Keep it in the spec and on the Learn site, and maybe have a section on the website that states our position, because it's historical, legacy and critical. When we have a test suite, a conformance suite, that'll probably be its own site and repo, and we link them together: the spec says what we're doing, the website states our position and links to the spec and the conformance suite.

00:23:59Speaker [?]: That sounds feasible.

00:24:00Kin Lane: When Joost is back, I see him helping lead the spec, and Dimitri and Joost bringing vacuum alignment to the table. We're talking about several things: the OpenLint spec, vacuum's ruleset format, and Spectral's, which I'd say isn't a spec but a config. How do we acknowledge the alignment of these? Then there are tools like Redocly and others that import rulesets; they don't support it as a spec. I'm still going to ask them whether they're interested in supporting this spec, which for me means an import and an export. I'd approach Postman again too, since I helped build it into the Postman engine, though from working there I know they're probably not interested in exporting it. So the initial spec alignment work has to happen, then a conformance suite around it that says how each tool conforms: OpenLint against Spectral, vacuum and OpenLint, and how Spectral, vacuum and others conform. Some sort of matrix, so people can see in one glance how it works.

00:26:39Nangiale Shir: Sorry, I didn't want to interrupt. In addition to what Dimitri said: I'm an engineer from Germany, working for the Open Resource Discovery organization, where we currently use Spectral for the validator. Pavel invited me to this call. Spectral is a core tool in our validator. We built a lot of rules and rulesets around it, but the core is based on Spectral. Like Dimitri, we have pipelines with large data, and there's a need for optimization, with potential on the Spectral side. We're really happy that a fork like OpenLint is established. We're observing it, and maybe we'll migrate if the community is active and feedback is received and responded to.

00:28:35Kin Lane: Thank you for jumping in.

00:28:39Nangiale Shir: It's my first call, and it's more observing how it's built. It's not urgent for us to migrate directly. I've already contributed to OpenLint, with a performance optimization for reusing validation schemas on the AJV side of Spectral. There are a lot of other optimization opportunities too, so I think it's a good idea to join the call in the future and see how it grows, and when it's ready for us to migrate. It will happen.

00:29:53Kin Lane: Thank you for joining. Conversations with Pavel back in July were one of the things that pushed me to fork this, along with IBM, Microsoft and others saying they'd very much like this to move forward. The goal has been to isolate the spec as a spec, because it wasn't one, which we've done; stabilize the tool; deal with the supply chain issues and the telemetry concerns; and get an npm package up and working as is. We finally achieved that last week, and now we're starting to consider the roadmap. I'll take in your performance considerations as part of that. Making it faster and better without breaking anything is ideal. But as we just discussed, we have a lot of backwards compatibility, and horizontal compatibility with vacuum and others, to think about, and we need a conformance suite to make sure we're doing it right. We have a lot of trust to establish, and seeing the Dutch government involved is part of the storytelling. We'll officially launch at API Days Paris in December.

00:32:19Nangiale Shir: You mentioned it will be officially released in Paris?

00:32:26Kin Lane: It's released now. The theatrical public release will be at API Days Paris. We consider the package stable enough: I've switched out my infrastructure, the Dutch government has, and we're encouraging others to run pilot projects. If you want to be part of the announcement, there will be several talks: the Dutch government, me, Andrzej I think, and others. The second part of API Days Paris brings together the European Commission and government agencies, and I'll give another talk there. There'll be a press release. It'll be a theatrical moment, but it's already there.

00:33:46Nangiale Shir: For the bigger audience, OK. I see that Spectral isn't being invested in as much any more, and now new rules come for MCP, A2A agent cards and agent skills. We already have those, specified for our needs under SAP's strict guidance, built around Spectral as the core. I'll report this back and see how we can move forward.

00:35:26Kin Lane: A couple of things. SmartBear has started doing more work since we forked. I know SmartBear; they'll follow our work and try to compete with us. They had made motions that they'd contribute Spectral to this group, which didn't happen, and I'm not surprised. This isn't my first rodeo with SmartBear and specs. We'll keep paying attention to them and try to include them; they're always welcome to join these meetings. I wish them the best, but I'd rather they joined us. We'll take it week by week. On your support for MCP, A2A and others: when Pavel and I first talked in July, you were forking it in your own way, IBM was forking it and using it in their own way, and I'd forked it and started adding my own MCP and agent skills. Seeing everyone move it forward in their own way is what convinced me to do this. So I'd encourage you, and IBM is already doing this, to bring your needs to the table. MCP, A2A and these new formats should be your specs; you should define them and how they're used. I'm bringing my formats, like APIs.json, and I'd like to see SAP bring yours, Open Resource Discovery and others. That's an incentive to swap parts of the spec into your tools, so you can bring your needs to the table and help shape the roadmap.

00:38:01Nangiale Shir: For us, support for new document formats would be helpful, but we've already built around that. More important is how fast you review my performance contribution: an optional flag that would improve our pipelines and reduce workflow time. What matters to us is how fast it's reviewed, discussed and maybe merged. The fork of Spectral already covers our needs, because we have our own strict rules. It's more about performance, and whether the community behind it is fast. If it gets done, we're likely to get involved directly.

00:39:46Kin Lane: I hear you loud and clear. As I said in my comment on the pull request, we're just finishing setting the table: everything works, and a couple of core groups are swapping packages. I'd like to add it to the roadmap and start prioritizing it. Give me some time to review it, and I'll respond as rapidly as possible, to demonstrate that we'll move fast, but thoughtfully, with our roadmap.

00:40:29Kin Lane: So that's the rules architecture, a big one to set the table. I should also add a new roadmap section for performance, optimization and other housekeeping, because there's clearly a lot more work to do there. I'll add it after this meeting. Moving to the toolbox itself: Phil and Jakub aren't here, and Miguel has left, so I don't have much to add on the tools. I'm still working on the core ruleset based on my work. Ivan submitted a VS Code extension that we're talking about bringing into the org, and I set up a base prototype for an LSP server, so we can support VS Code but also other IDEs. Feel free to chime in. Miguel talked about the MCP server, an MCP server for OpenLint you can use in your LLM client of choice to get rules and submit them; we'd love more feedback on it. We have several other bugs and items, which I'll organize as housekeeping.

00:42:25Kin Lane: Joost isn't here. We have pull requests with updates to the spec, but we have a spec, and formal docs. There's housekeeping and quite a few issues to get through, which can happen asynchronously. I'll work with Joost and Dimitri on bringing in vacuum, Spectral, Redocly and other tooling support, as we just discussed. And I have to move the conformance suite forward, because I need to say that we're conformant with these other tools, and what they're conformant with. I can't just make it up every week; I need something to run, and a visual dashboard. Then we can address backwards compatibility. I need all this in place before we add anything new, even utility, housekeeping or performance, so we're not breaking anything for existing Spectral and vacuum users. And I still haven't been able to clearly articulate where vacuum deviated. I know vacuum added a couple of properties.

00:43:50Dimitri van Hees: I dived into that. Perhaps I can open a new discussion about it. There are some extra features we could add to the current spec. For example, extra properties like category, which isn't possible in Spectral, but it appears it is possible with an x- prefix. Those are missing features that would be very handy for us. But the main issue is the programming-language dependency. vacuum uses regular expressions and plain JSONPath, while Spectral uses JSONPath Plus, which is a JavaScript library. So vacuum's take on the format is, in my opinion, from a standardization point of view, better than Spectral's. I'll write it all down.

00:45:13Kin Lane: Once you write that down, I'd like to create example rules, because as I set up the conformance suite, we need rules that show these differences.

00:45:28Dimitri van Hees: I can make a matrix of all the differences between vacuum and Spectral, and then we can say we want this in OpenLint as well, and use that as input for the next version of the spec. That's basically what I meant with the discussion I started, but I want to look further into it.

00:45:50Kin Lane: I think the x- path is how we maintain backwards compatibility while slowly introducing new properties. Eventually we'll have to go down a breaking-change path from Spectral, but maybe we can get them to update their tools. My goal is to eventually get SmartBear to support OpenLint, and of course vacuum.

00:46:24Dimitri van Hees: I have to run, but one more update worth mentioning. The standardization board wants to address this in next year's budget, because they're looking for maintainers. The question is always who maintains this. For the OpenAPI standard it's the Linux Foundation, and they trust them. But perhaps they'll want the Dutch government to help maintain it. They want to investigate how they can contribute and support it. Probably not by paying money directly to Open Source Europe, because that's difficult in government finance. Joost and I are doing this in another project's time, but perhaps they'll free up budget to dedicate people to maintaining it, or helping to.

00:47:32Kin Lane: Thank you. Let us know what else we can do to support that, including shifts in governance strategy that make the government feel better. That's a nice segue into administration. Now that we have our fiscal partner, I'm crafting a handful of grant proposals around the conformance suite, the spec work, the Learn section, and the rulesets, including industry rulesets, to find a sweet spot for chasing European money. Open Source Europe as fiscal sponsor is one thing, but if we need more of a corporate or European entity, maybe we can do something within the Dutch government, as Dimitri said. I'm open to discussion. There's a huge opportunity for this in Europe right now, as a governance standard working across specs, but we have a lot of trust to build and work ahead of us. That's the agenda. Anything else people want to raise?

00:49:03Kin Lane: This week I'm doing more work on rulesets and some work on the conformance suite. I may get time for the VS Code extension and the language server, or that might be what gets dropped. I'm letting Joost lead the spec, and Phil, Jakub and Miguel lead the toolbox, while I stay focused on rules, rulesets, learning, funding, and now the conformance suite, which I think is critical to building trust. I'll go through all the open items, bugs and the performance work. I think we're ready for an official roadmap, and we need a weekly approach to iterating on it. That came out of this conversation: it's one way we build trust. I was leaning heavily toward new formats, AI features and experience features, based on the conversations I'm having. But as Dimitri said, simply being maintained and having a community starts building trust, and responding to and iterating on the roadmap in an organized way is how we build more and get more people involved. Heard loud and clear.

00:50:59Kin Lane: You know where to find me. I'll keep building toward API Days Paris, but we're ready now for integration and design partners, and I'll be reaching out to more contacts to get people to switch. The official launch will be at API Days Paris, and we'll make a big to-do about it. Enjoy your weekend.