Claude Partner Network: 3 Tiers, and Where Agencies Should Certify

The Claude Partner Network now has a tier ladder, and it is built around proof of deployment rather than logos on a slide. On June 3, 2026, Anthropic announced the Services Track and a Partner Hub that ranks agencies by how many certified people they employ and how many live Claude deployments they can name. For agencies deciding where to invest, the practical question is not whether to join but which tier is realistically within reach this year.

The network itself launched earlier, on March 12, 2026, backed by what Anthropic describes as a $100 million ecosystem investment for partner training, technical support, and shared marketing. Joining is free and open to any organization bringing Claude to market. The June announcement is what turned an open network into a structured program with named tiers, fixed promotion dates, and a public directory that enterprise buyers can search.

This guide covers what actually launched and when, the exact requirements for each of the three tiers, the certification exam that underpins them, and the part no press coverage has done cleanly: the arithmetic of reaching Select tier as a small agency, and how the Claude program compares with the partner ladders at AWS, Salesforce, and OpenAI.

It helps to separate the two events. The Claude Partner Network launched on March 12, 2026 as an open network: free to join, open to any organization bringing Claude to market, and accompanied by the first official certification. That launch came with the headline financial commitment — a $100 million investment that Anthropic says is allocated to partner training, technical support, and shared marketing across 2026.

The Services Track and Partner Hub, announced June 3, 2026, are the structure layered on top. This is where the three tiers, the fixed promotion calendar, the daily-refresh standing dashboard, and the public buyer directory all live. Anthropic reports that over 40,000 firms have applied to join the network since March, and that more than 10,000 consultants have earned Claude certifications through the Partner Academy. Both figures are Anthropic's own; no independent audit has been published, so treat them as the company describes them rather than as verified market data.

Each tier is defined by two countable things: the number of active certified individuals on staff, and the number of joint customers in production. The thresholds escalate sharply, which is what makes the ladder a useful filter — the requirement structure tells you who each tier is built for. A point worth internalizing early: certifications are held by individuals, not by the firm. A firm's tier standing is the count of its active certified employees plus its cumulative production deployments, so headcount turnover directly affects standing.

The promotion calendar matters as much as the thresholds. Anthropic sets tier promotions on January 1 and July 1 each year, with an additional review on October 1, 2026 that applies in the first year only — not a permanent third window. There is an annual review on December 31. The interpretation worth drawing out: any agency that clears Select criteria before the October 1, 2026 review gets listed in that cycle, ahead of the larger January cohort. For a small firm, that first-mover window is the difference between appearing in a sparse directory and appearing in a crowded one.

One more structural detail that rewards close reading: business development and practice building are tracked separately. Partners earn credit in two distinct categories rather than against a single blended score, which means an agency strong on delivery but light on co-sell can still build standing on the practice-building side. That separation is unusually friendly to delivery-first boutiques.

Every tier threshold counts certified individuals, so the exam is the unit of progress. The Claude Certified Architect (CCA) — Foundations is the first official Anthropic certification, launched alongside the March network announcement. It is a proctored, 60-question exam delivered on the Anthropic Academy platform, runs 120 minutes, and is scored on a 1,000-point scale. The standard cost is $99, and the first 5,000 partner employees can sit it for free — an early-access window that materially changes the certification math for firms that move now.

The reported passing threshold is 720 out of 1,000, according to multiple third-party exam-preparation resources. Anthropic has not published this score on its primary announcement page, so treat it as community-reported rather than confirmed policy and verify the current standard on the Partner Academy before planning a certification sprint.

The weighting tells you what Anthropic considers foundational for a services partner: nearly half the exam (47%) sits in agentic architecture plus Claude Code workflows, with prompt engineering, tool and MCP design, and reliability filling the rest. For an agency planning who to certify first, the practical read is to start with the engineers and solution architects who already build agent systems — they cover the heaviest-weighted domains by experience rather than study.

Anthropic has also signaled that more credentials are coming. Seller and developer certifications plus an Advanced Architect tier are planned for release later in 2026, and the company has stated that industry-specific specializations are on the Services Track roadmap. The specializations remain an announced intent rather than a published program — no specific verticals or launch dates have been confirmed — so plan around Foundations today and treat the rest as forward signal.

Here is the core of why Select is the boutique play, laid out as arithmetic that no other coverage has run. Select requires 10 active certified individuals. At the standard $99 exam fee, certifying 10 people is roughly $990 in direct exam cost — and if an agency claims seats inside the first-5,000 free window, the direct exam cost can be zero. Even at full price, a 15-person agency can put its whole technical bench through CCA-Foundations for under $1,500 in exam fees.

The genuinely expensive part is not the exam. It is the second half of the requirement: 2 joint customers in production over the trailing 12 months, plus 1 public customer story. Production deployments take real delivery time, and a public story takes client permission. That is the constraint a small agency should plan around, not the certification fee.

That line points to the single highest-leverage process change a Select-bound agency can make: build customer-story permission into the engagement contract upfront. Asking a client for case-study rights after a project closes is materially harder than securing them as a standard clause at kickoff. An agency that adds a permission clause to its standard statement of work today removes the slowest dependency on the Select path before it even starts certifying staff.

Tiered partner programs are not new — AWS and Salesforce have run them for years — so the useful question is how Claude's thresholds compare, and where its model diverges. The table below puts four programs on a single grid. The pattern it reveals: Anthropic requires relatively few certified staff at entry (10) but demands explicit production-deployment proof at every tier, while OpenAI has no open tiered structure at all.

The strategic contrast is OpenAI's. In February 2026 OpenAI launched Frontier Alliances with BCG, McKinsey, Accenture, and Capgemini — a non-tiered, named-partner model rather than an open ecosystem. OpenAI does run a certification program, so it is not accurate to say it has no partner motion. But there is no public application pathway equivalent to the Claude Partner Network for a mid-market agency, which concentrates OpenAI's services ecosystem at the top while Anthropic's open-tier model favors ecosystem depth. For a boutique firm that wants a published path to enterprise visibility, that difference is decisive — and it is the same divide we unpack in our look at OpenAI's Frontier Alliance structure.

Salesforce's own 2026 overhaul reinforces the direction. It collapsed four tiers into two and folded 170 badges into 28 competencies tied to Agentforce and Data Cloud — a deliberate move away from breadth-of-logo toward demonstrable, deployment-anchored depth. Read across all four programs, the cross-vendor signal is consistent: the era of partner status earned by attending trainings is ending, and proof of production deployment is becoming the currency everywhere.

Most coverage frames the Services Track from the supply side — the effort an agency expends to climb. The underreported half is the demand side. The Partner Hub doubles as a public enterprise buyer directory, surfacing qualified Claude implementation partners by certified headcount and deployment history. Any agency that reaches Select becomes visible to enterprise buyers browsing for Claude partners. That reframes the whole exercise: certification is not compliance, it is lead generation.

The Hub also includes a standing dashboard that refreshes daily, showing each partner's live position against its target tier's requirements — certified headcount, deployed customer count, and public story count. Anthropic has embedded a Model Context Protocol connector so partners can query that standing conversationally through Claude itself, asking where the firm stands against, say, Preferred-tier requirements. It is a small but telling detail: Anthropic is using Claude to run the workflow that manages the Claude partner program.

The enterprise demand is not hypothetical. Anthropic announced large deployments alongside the Services Track: it states Accenture is training 30,000 professionals on Claude, Cognizant has rolled it out to roughly 350,000 associates, Deloitte is making it available to 470,000 people globally, and KPMG is integrating it across a workforce of more than 276,000. These are Anthropic-stated figures rather than independently audited adoption, but the direction is unambiguous: the buyers searching the Partner Hub are some of the largest professional- services firms in the world. The depth of that enterprise push is the same one we cover in our look at Accenture's Claude partnership and Anthropic's broader security and enterprise expansion.

The tier thresholds are the same for everyone, but the realistic 12-month target is not. The table below maps agency size to a sensible tier ambition, the certified headcount and deployments that implies, and the binding constraint for each band. It is a planning grid, not a guarantee — deployment timelines vary by client and engagement type.

The grid makes the central recommendation visible. For the bulk of independent agencies — the 3-to-50-person band — Select is the target that is both meaningful and attainable inside a year, and the binding constraint is almost always the customer-story permission, not the certification fee or even the deployment count. Solo operators should join the network now to start the relationship and accumulate deployment evidence, even though the 10-person floor puts Select out of reach until they grow. The decision of whether certification ROI justifies the delivery time is exactly the kind of question our guide to pricing Claude-powered agency services is built to answer.

For a boutique agency that wants to reach Select before the October 1, 2026 review, the sequence matters more than the effort. The four moves below are ordered by dependency: the slowest item (story permission) starts first, the cheapest item (certification) can run in parallel.

The thread running through all four moves is that proof of deployment, not certification volume, is the real work. The exam is cheap and fast; the production deployments and the public story are where the quarter goes. An agency that internalizes this — and front-loads the permission clause and the deployment documentation — can reach Select with a small, deliberate team. If you want a second set of eyes on whether the certification investment pencils out against your delivery capacity, our AI transformation engagements start with exactly that kind of assessment.