Anthropic Cowork Pricing Cost Free Tier Firm Procurement

How much does Anthropic Cowork cost? The plugin itself is free; the Claude subscription underneath isn't. Per Anthropic's official pricing page as of April 28, 2026, the relevant tiers stack from $0 (Free tier) up to $125/seat/month (Team Premium monthly) with Enterprise at $20/seat/month + usage at API rates for custom-terms deployments. The Cowork legal plugin is open source and free per the GitHub repo. This is the head-on cost analysis: every tier, what's included, what it costs annually for typical firm sizes, and where the procurement decision actually sits.

The full Anthropic pricing table — every tier, every rate

Per Anthropic's pricing on April 28, 2026:

Consumer tiers: - Claude Free — $0/month. Basic chat access on web, iOS, Android, and desktop. Plugin install possible but matter work shouldn't run here. - Claude Pro — $17/user/month annual or $20/user/month monthly. Includes Claude Code and Cowork. Floor for individual lawyer experimentation. - Claude Max — $100/user/month starting price (monthly). 5x or 20x more usage than Pro depending on tier. For power users who hit Pro's usage caps.

Team tiers (firm deployments with admin controls): - Claude Team Standard — $20/seat/month annual or $25/seat/month monthly for 5-150 seats. Floor for firm deployments. Admin controls plus explicit no-training commitment on team inputs. - Claude Team Premium — $100/seat/month annual or $125/seat/month monthly. Premium seat tier with higher usage caps. Right for power users running multi-day diligence work.

Enterprise tier: - Claude Enterprise — $20/seat/month + usage at API rates with custom terms. Advanced security, compliance, custom data residency. For firms negotiating bespoke compliance requirements.

API pricing (token-based): - Opus 4.7 — $5/M input tokens + $25/M output tokens. Most capable model. Batch processing 50% off; US-only inference 1.1x. - Sonnet 4.6 — $3/M input tokens + $15/M output tokens. Mid-tier model. Most cost-effective for routine work. - Haiku 4.5 — $1/M input tokens + $5/M output tokens. Fastest, cheapest tier.

The Cowork legal plugin is free and open source per the GitHub repo — the plugin itself adds no cost on top of the Claude subscription.

Annual cost by firm size — three deployment scenarios

Three typical deployment scenarios with full annual math:

Solo practitioner or sub-10-lawyer firm: - Claude Pro at $17/user/month annual = $204/year per lawyer - Cowork legal plugin: $0 - Configuration time (one-time): 4-8 hours of senior counsel time - Total annual: roughly $200-2,000/year depending on firm size

Mid-market firm — 50-lawyer in-house legal department: - Claude Team Standard at $20/seat/month annual = $12,000/year for 50 seats - Cowork legal plugin: $0 - Configuration time (one-time): 8-16 hours legal-ops plus 4-8 hours senior counsel review - Total annual: $12,000-15,000/year all-in

Mid-market firm — 200-lawyer mixed practice: - 80% Claude Team Standard (160 seats × $20/seat/month annual) = $38,400/year - 15% Claude Team Premium (30 seats × $100/seat/month annual) = $36,000/year - 5% API allocation for internal tooling builds = $15,000-30,000/year depending on usage - Cowork legal plugin: $0 - Total annual: $89,400-104,400/year for the model layer

BigLaw — 1,000-lawyer firm on Enterprise tier: - Claude Enterprise at $20/seat/month + usage rates = $240,000/year base for 1,000 seats - Plus usage at API rates (Opus 4.7, Sonnet 4.6, Haiku 4.5 mix) - Plus deployment surface fees (Microsoft Foundry, AWS Bedrock, or Vertex AI inheritance) — depends on existing cloud relationship - Custom terms negotiated with Anthropic - Total annual: typically $500,000-1,500,000+ depending on usage and custom terms

The second-order read: Cowork plugin's free status means the procurement battle is at the Claude subscription tier, not at the plugin. Firms already paying for Claude get the plugin at no incremental cost. Firms not on Claude have a procurement question about which tier to start with.

Free tier — what you get and why it's not appropriate for matter work

Claude Free at $0/month provides basic chat access. The Cowork legal plugin can technically install on a Free account. Don't use it for matter work. Per Anthropic's terms, Free and Pro tier inputs may be used for model training. Privileged or commercially sensitive matter content cannot run on these tiers.

What Free tier is appropriate for:

- Personal experimentation with the plugin commands on synthetic or public data. - Learning the workflow before purchasing Pro or Team. - Demonstrating value to colleagues using non-confidential samples.

What Free tier is NOT appropriate for:

- Any privileged matter - Any client-confidential content - Any work product that would normally fall under attorney work-product doctrine - Any production legal workflow

Per US v. Heppner (SDNY, February 17, 2026), consumer Claude communications were not protected by attorney-client privilege or work-product doctrine. The Heppner ruling specifically addressed consumer Claude (which includes Free and Pro tiers). Enterprise tiers (Team, Enterprise, API) carry stronger data-handling commitments and a different privilege analysis. The Heppner enterprise privilege defense stack spoke covers the four-layer architecture.

Pro tier — when $17-20/user/month makes sense

Claude Pro at $17/user/month annual or $20/user/month monthly is the floor for individual lawyer use. Per Anthropic's pricing, Pro includes Claude Code and Cowork plus the Cowork legal plugin once installed.

Where Pro tier fits:

- Solo practitioners building their own AI workflows. $204/year all-in for individual deployment is on a personal credit card budget. No procurement involvement needed. - Sub-10-lawyer firms experimenting with AI. Each lawyer can deploy Pro on their own account and experiment with the plugin against synthetic data before deciding on firm-wide procurement. - Power users at larger firms supplementing their firm's Team deployment. When usage caps hit, Pro adds capacity outside the Team contract.

Where Pro tier doesn't fit:

- Firm-managed deployments where matter confidentiality matters. Pro tier inputs may be used for training. Matter work needs Team or higher. - Privileged or commercially sensitive content. Pro tier doesn't carry the no-training commitment Team and Enterprise do. - Multi-user firm deployments. No admin controls, no centralized policy enforcement.

The second-order read: solo practitioners are the cleanest fit for Pro tier. Per the bottom-up procurement model analysis, individual Pro adoption at firms produces the demand signal that drives firm-wide procurement to Team or Enterprise. The Pro stage is short-lived for firms that move to deployment.

The third-order read: Anthropic's pricing structure is designed so that Pro experimentation flows naturally into Team procurement. The economics encourage the upgrade path the firm needs anyway.

Team and Enterprise — the firm-deployment tiers

Claude Team Standard at $20/seat/month annual or $25/seat/month monthly for 5-150 seats is the floor for firm deployments. Per Anthropic's pricing, Team Standard includes:

- Admin controls for centralized policy enforcement - Explicit no-training commitment on team inputs - Centralized billing and seat management - Standard data-handling guarantees Anthropic doesn't extend to consumer tiers

Claude Team Premium at $100/seat/month annual or $125/seat/month monthly is the right tier for power users:

- 5x higher usage caps than Team Standard - Same admin controls and no-training commitment - Right for M&A diligence partners running multi-day multi-session memory workflows - Right for litigation partners running heavy discovery review

Claude Enterprise at $20/seat/month + usage at API rates with custom terms is for:

- Firms with 500+ headcount needing custom data residency, advanced security, compliance certifications - Firms negotiating bespoke contractual terms (data processing agreements, SLAs, audit rights) - Firms requiring deployment via Microsoft Foundry, AWS Bedrock, or Vertex AI per the Microsoft Foundry deployment analysis - Firms with regulated-industry practice areas needing specific compliance certifications

The second-order read: most mid-market firms (50-500 lawyers) need Team Standard plus selective Premium upgrades. Enterprise tier matters at 500+ headcount or for firms with bespoke compliance requirements that standard Team tier doesn't address.

The third-order read: Enterprise tier negotiation is where the largest firm-tier discounts surface. List prices are starting points; multi-year commitments and large-seat counts produce material discounts in negotiation. Per the procurement checklist for mid-market firms, firms negotiating Enterprise should expect 3-6 month sales cycles with custom term sheets.

Cost comparison — Cowork vs alternatives at typical firm size

Two head-to-head pricing comparisons at the 200-lawyer mid-market firm size:

Cowork plugin on Claude Team vs Spellbook: - Cowork plugin on Claude Team Standard: $48,000/year (200 seats × $20/seat/month annual). Plugin free. - Spellbook: industry estimates of $180-300/seat/month per Artificial Lawyer and aiapps coverage = $432,000-720,000/year for 200 seats — NOT vendor-confirmed. Per Spellbook's pricing page, all tiers are quote-only. - Premium for Spellbook: roughly 9-15x Cowork on Claude Team Standard, paying for vertical contract review specialization plus the CBA exclusive partnership.

Cowork plugin on Claude Team vs Microsoft Copilot for legal: - Cowork plugin on Claude Team Standard: $48,000/year for 200 seats. - Microsoft 365 E3 + Copilot: $36 (E3) + $30 (Copilot) = $66/user/month annual = $158,400/year for 200 seats per Microsoft's pricing. - Premium for Copilot: roughly 3.3x Cowork on Claude Team Standard, paying for in-suite integration into Word, Outlook, Teams, Excel. - Most firms running Microsoft 365 already pay the E3 component anyway, so the marginal Copilot add-on is $30/user/month or $72,000/year for 200 seats.

Cowork plugin on Claude Team vs Thomson Reuters CoCounsel: - Cowork plugin on Claude Team Standard: $48,000/year. - TR CoCounsel: industry estimates of $75 (On Demand) up to $500 (All Access) per user/month per Costbench March 2026 and Above the Law August 2025 coverage — NOT vendor-confirmed. For 200 seats: $180,000-1,200,000/year depending on tier. - Premium for CoCounsel: 4-25x Cowork on Claude Team Standard, paying for Westlaw and Practical Law content embedded in workflow.

The second-order read: Cowork plugin on Claude Team is the cheapest deployment for the configurable workflows it covers. The premium for vertical vendors (Spellbook, CoCounsel) pays for proprietary content depth and vertical specialization that Cowork doesn't replicate. The premium for Copilot pays for productivity-suite integration. Most mid-market firms running this comparison correctly end up choosing combinations rather than substituting.

The third-order read: per the Cowork vs Microsoft Copilot vs Spellbook vendor war analysis, the right answer for 200-lawyer mid-market firms is typically Cowork plus Copilot plus selective Spellbook for transactional-heavy practice areas. Total stack annual lands at $90,000-208,000/year depending on the practice mix.

The Bottom Line: My take: Anthropic Cowork pricing is straightforward — Free tier ($0, not for matter work), Pro ($17-20/user/month, individual), Team Standard ($20-25/seat/month, firm floor), Team Premium ($100-125/seat/month, power users), Enterprise ($20/seat + usage rates, custom terms). The Cowork legal plugin itself is free and open source. For a 200-lawyer mid-market firm, the model layer alone lands at $90,000-108,000/year. Cowork plugin on Claude Team Standard is structurally cheaper than Spellbook (9-15x premium for vertical specialization), Microsoft Copilot (3.3x premium for productivity-suite integration), and TR CoCounsel (4-25x premium for proprietary research content). The procurement decision isn't 'which is cheapest' — it's 'which combination fits the firm's practice mix and existing infrastructure.' Most firms running this analysis correctly end up combining Cowork with one or two adjacent tools rather than substituting.