OpenAI on Tuesday released six role-specific plugins for Codex, its coding agent turned enterprise productivity platform, targeting public equity investing, investment banking, sales, data analytics, product design, and creative production. The plugins bundle 62 third-party apps and 110 discrete skills, according to OpenAI’s announcement. Legal, corporate finance, private equity, marketing strategy, and strategy consulting plugins are coming next.

The timing is deliberate. Anthropic filed its S-1 with the SEC on June 1, seeking a reported $965 billion valuation. OpenAI’s own IPO filing is expected within weeks, at an $852 billion valuation. Both companies need to prove that enterprise agent revenue is real, scalable, and defensible. The plugin launches are their respective pitches to investors.

What OpenAI Actually Shipped

Six plugins, each targeting a specific job function:

The public equity investing plugin connects to Moody’s, Daloopa, Datasite, FactSet, LSEG, S&P, PitchBook, and Hebbia. Investors can review earnings, compare companies, track signals, and stress-test investment theses against live financial data, per OpenAI.

The investment banking plugin handles pitch materials, comparable company analysis, transaction comparables, and diligence-to-recommendation workflows. OpenAI describes the data sources as “trusted” but did not name specific providers beyond the equity investing stack.

The sales plugin integrates Salesforce, HubSpot, Slack, Outreach, Clay, Rox, and Actively. It identifies high-priority accounts, prepares meeting briefs, manages follow-ups, and flags at-risk deals, according to SiliconANGLE.

The data analytics plugin supports Snowflake, Databricks Genie, Hex, and Tableau. Teams ask business questions in plain language and get reports, dashboards, and metric explanations back.

The remaining two plugins cover product design (prototyping from URLs, user flow audits, Figma and Canva integration) and creative production (campaign boards, display ad variations, product lifestyle shots via Figma, Canva, Shutterstock, Picsart, and Fal).

Alongside the plugins, OpenAI launched Sites, a feature that lets Business and Enterprise customers create interactive, hosted web pages from Codex outputs, shareable via URL. Partners including Vercel, Wix, Replit, Lovable, Figma, and Webflow are part of the early ecosystem. OpenAI also extended its Annotations feature, previously limited to code, to non-technical documents: users can highlight a section of a spreadsheet or slide, describe what should change, and Codex modifies only that region. SiliconANGLE reported this positions Codex as an iterative refinement tool, not just a first-draft generator.

The Growth Numbers Behind the Pivot

Codex now has more than 5 million weekly active users, a sixfold increase since the desktop app launched in February 2026, according to OpenAI. Non-developers, including analysts, marketers, designers, investors, and bankers, make up 20% of that user base and are growing more than 3x faster than developers.

That growth rate matters because it signals where the revenue is heading. PYMNTS reported that OpenAI generated nearly $6 billion in first-quarter 2026 revenue, driven partly by Codex, business sales, and early ChatGPT advertising tests. Enterprise now accounts for more than 40% of total revenue, per analyst estimates, on pace to reach half by year-end.

OpenAI’s annualized revenue has climbed from $2 billion in 2023 to roughly $25 billion by early 2026. For the IPO, the question is whether Codex’s non-developer expansion is a product-led growth engine or a one-time adoption wave that flattens once the novelty fades.

Anthropic’s Opposite Bet: Go Deep, Not Wide

While OpenAI launched six plugins across six job functions, Anthropic has spent 2026 doing the opposite: going deep into a single vertical.

Claude for Legal, which Anthropic expanded aggressively in May, now includes 12 practice-area plugins covering M&A due diligence, employment handbook drafting, litigation support, and compliance workflows. The platform has more than 20 connectors to legal-specific tools including Westlaw, CourtListener, and iManage. Anthropic also shipped a cross-app Microsoft 365 integration that embeds Claude across Word, Outlook, Excel, and PowerPoint as a single context-carrying agent.

The legal industry is responding. Freshfields, Quinn Emanuel Urquhart & Sullivan, Holland & Knight, and Crosby Legal have all adopted Claude for live matters, Fortune reported. More than 20,000 legal professionals attended Anthropic’s most recent legal webinar, its largest legal session ever. Claude Opus 4.7 scored 90.9% on Harvey’s BigLaw Bench, the legal industry’s primary AI benchmark.

Christopher Kercher, partner and head of AI at Quinn Emanuel, told Fortune he built the firm’s litigation platform on Claude “with virtually no coding background” because he needed it for an actual trial. “The work product is far beyond what I would’ve done on my own, probably ever,” he said.

Anthropic’s legal strategy has a distinct architectural advantage over OpenAI’s approach: grounding. Rather than generating answers from training data, Claude for Legal’s connector architecture forces the model to draw only from verified sources, including actual court opinions, document repositories, and case law databases. “In litigation, an authoritative-sounding hallucination is worse than no answer,” Jay Madheswaran, CEO of legal AI company Eve, told Fortune.

LawSites reported Anthropic released all of its legal plugins as open-source on GitHub, available both as Cowork plugins and through the Claude Managed Agents API. That dual-distribution model lets law firms install plugins directly in Claude’s desktop interface or deploy them behind their own workflow engines.

Horizontal vs. Vertical: The Strategic Tradeoff

The two approaches represent fundamentally different theories about how enterprise agent revenue scales.

OpenAI is betting on breadth. Six plugins launched simultaneously, covering finance, banking, sales, analytics, and design, with legal, corporate finance, and consulting coming next. The thesis: Codex becomes the universal agent platform, the workspace where every knowledge worker starts their day. Revenue scales through volume, as 5 million users become 50 million, and through upselling, as free ChatGPT users convert to Business and Enterprise subscriptions to access plugins and Sites.

Anthropic is betting on depth. One vertical, saturated with connectors, grounded in domain-specific data sources, validated by industry benchmarks, and adopted by named firms with reputations on the line. The thesis: a model that scores 90.9% on BigLaw Bench and connects to Westlaw is worth more per seat than a general-purpose agent with Salesforce integration. Revenue scales through high-value contracts with regulated industries that pay premium rates for reliability.

The Decoder noted that Anthropic has been offering similar plugins since January 2026, describing them as “a mix of agent prompts (‘skills’), predefined tools like web search, and data connectors.” OpenAI’s plugin launch comes five months later, per SiliconANGLE.

That five-month gap is telling. Anthropic chose to iterate in legal specifically, refining connectors and building credibility with AmLaw 200 firms. OpenAI chose to wait and launch wide, covering multiple job functions simultaneously but with less vertical depth in any one of them. OpenAI’s equity investing plugin, for example, connects to eight financial data providers. Anthropic’s legal plugin connects to more than 20 legal-specific platforms and includes 12 practice-area workflows.

The IPO Pressure

Both companies are pitching enterprise agent revenue to public market investors at roughly the same moment.

OpenAI’s numbers favor the horizontal story: 5 million weekly Codex users, 3x non-developer growth rate, $25 billion annualized revenue, and 40%+ enterprise revenue share. The plugin expansion is designed to show that Codex can reach every knowledge worker, not just the 80% who write code.

Anthropic’s numbers favor the vertical story: named law firm adoptions, a benchmark score investors can point to, and an open-source distribution model that positions Claude as infrastructure rather than a consumer product. The New York Times reported that Anthropic’s explosive growth over the last year came “largely thanks to technology that can automatically write computer code,” suggesting its own coding revenue is substantial but that legal represents the highest-margin expansion opportunity.

The question for investors is whether enterprise agents are a winner-take-most market (favoring OpenAI’s breadth) or a verticalized one where domain expertise creates durable moats (favoring Anthropic’s depth). The answer probably depends on the industry. Financial services, where OpenAI just launched two plugins, is notoriously sensitive to errors in numerical analysis. Legal, where Anthropic has invested most heavily, punishes hallucination with sanctions and malpractice claims. Both verticals demand the kind of grounded, verified outputs that general-purpose models historically struggle to deliver.

The Open Ecosystem Question

OpenAI’s announcement included one line that may matter more than the plugins themselves: “We’re building toward an open ecosystem where partners can create and deploy their own plugins directly in Codex and ChatGPT.”

If that ecosystem materializes, the competitive dynamic shifts. OpenAI would no longer need to build every vertical plugin in-house. Domain experts in healthcare, insurance, accounting, and construction could build their own Codex plugins the way developers build apps for the App Store. The plugin directory becomes a distribution channel, and OpenAI takes a platform cut.

Anthropic’s GitHub release of Claude for Legal suggests it sees the same future from the opposite direction: by open-sourcing its vertical plugins, Anthropic invites the legal community to fork, extend, and improve them, creating a community-maintained vertical that becomes harder to replicate than a proprietary plugin.

The race is no longer about which model is smarter. It is about which plugin ecosystem becomes the default workspace for knowledge workers in regulated industries. OpenAI is betting that breadth of integrations gets it there first. Anthropic is betting that depth of trust does.

The IPO filings will show which bet the market prefers.