Spellbook vs Claude Cowork Legal Plugin Contract Review

Two contract review paths sit at opposite ends of the build-vs-buy spectrum. Spellbook is the freshly-capitalized SMB-and-mid-market vendor, post-$50M Series B at $350M valuation, shipping a Microsoft Word add-in with quote-only seat licensing. Anthropic's Claude Cowork legal plugin is open-source, free, and structurally a build-your-own option that requires firm-side AI capability to deploy. Per Anthropic's plugin page and the GitHub repository, the Cowork legal plugin includes `/review-contract` (clause-by-clause analysis with GREEN/YELLOW/RED flags + redline suggestions) and `/triage-nda` (rapid pre-screening) commands. The vendor procurement vs in-house build decision is structural, not all firms have the internal capability to maintain an open-source legal AI deployment, and not all firms have procurement budgets that justify per-seat vendor licensing. Here's the operator read on which path fits which firm.

What each path actually delivers

Spellbook (vendor procurement path):

• Microsoft Word add-in deployment. IT effort: low. Time-to-deployment: typically days, not weeks.
• Purpose-built clause-by-clause contract review with playbook configuration.
• Spellbook says Library indexes the documents a firm connects or uploads for search from Word (spellbook.com/blog/library, read October 26, 2026); it does not say it trains a model on them.
• Vendor-managed model updates, integration improvements, and support.
• Per-seat licensing, quote-only (see the cost section below).
• 7-day free trial.
• For Canadian firms: CBA member preferred-access pricing per the March 2026 partnership.

Claude Cowork legal plugin (build-your-own path):

• Open-source, free. Released by Anthropic February 2026, see the Anthropic Legal Ecosystem map for the full strategic context.
• `/review-contract` command for clause-by-clause review against configured negotiation playbook. GREEN/YELLOW/RED flag pattern with redline suggestions.
• `/triage-nda` command for rapid pre-screening categorizing NDAs into standard approval / counsel review / full review.
• Configurable to org's playbook and risk tolerance.
• "Assistance not advice" disclaimer.
• Underlying model cost: Claude Opus 4.7 at $5/M input + $25/M output tokens, or Claude Pro/Team subscription at $20-$25/user/month.
• Deployment effort: requires firm-side technical capability to install, configure, integrate with internal systems, and maintain.

The second-order point: the Cowork plugin's $0 license cost obscures real total cost of ownership. Firms with internal AI capability and existing Anthropic enterprise relationships absorb the build-and-maintain cost as a fraction of existing engineering capacity. Firms without that capability face hiring costs, contractor costs, and ongoing maintenance overhead that exceeds Spellbook's per-seat licensing for typical mid-market deployments.

The third-order point: the Cowork plugin's market impact has been structural, not just product-level. Per Canadian Lawyer's coverage, the February 2026 Cowork launch coincided with significant market reactions: Thomson Reuters -16%, RELX -14%, Wolters Kluwer -13%, ~$285B market cap wiped in a single trading session. The strategic threat to legacy legal data vendors was the open-source-ness, not the model quality. That dynamic affects Spellbook's competitive positioning differently than it affects CoCounsel or Lexis.

Where Spellbook is the right path: vendor procurement fit profile

Spellbook's structural fit is firms without significant in-house AI engineering capability who want fast time-to-deployment on a contract review workflow:

The general-model route is tested in Claude Opus 4.7 against Spellbook for contract review.

• Solo and small firms (2-5 attorneys). Building and maintaining an in-house Cowork plugin deployment requires technical capacity these firms don't have. Spellbook's Word add-in is plug-and-play: install, configure playbook, deploy across the firm in days.
• Mid-market firms (10-50 attorneys) without dedicated IT or AI engineering staff. The total-cost-of-ownership math typically favors vendor procurement over hiring or contracting for ongoing maintenance.
• Firms with established Microsoft 365 procurement velocity but no Anthropic enterprise relationship. Spellbook's Word add-in deploys against existing M365 infrastructure. Cowork plugin requires Anthropic API integration and ongoing model-cost tracking.
• Canadian firms covered by the CBA exclusive partnership. Per the CBA partnership analysis, preferred-access pricing shifts the per-seat economics meaningfully and the procurement default shifted to Spellbook in March 2026.
• Firms wanting precedent learning out-of-the-box. Spellbook Library is purpose-built. The Cowork plugin doesn't have an equivalent native feature: implementing precedent learning on top of the open-source plugin is custom engineering work.

The fit signal: if procurement velocity, deployment ease, and predictable per-seat costs matter more than $0 license cost and customization depth, Spellbook is the structural path.

Where Cowork is the right path: build-your-own fit profile

Cowork's structural fit is firms with internal AI capability and customization needs that vendor products don't meet:

• AmLaw 100 firms with established AI engineering teams. Freshfields' Anthropic deployment (per the Freshfields × Anthropic analysis) is the precedent. Firms at this scale have engineering capacity to absorb Cowork plugin maintenance and benefit from full customization.
• In-house legal departments with internal IT and procurement velocity that doesn't require external vendor relationships. The Cowork plugin's open-source posture lets in-house teams deploy without going through firm-vendor procurement processes: IT can install it without procurement approval.
• Firms with non-standard contract types (energy sector, life sciences IP, multi-jurisdictional regulatory) where Spellbook may not fit. Custom configuration on the Cowork plugin lets the firm tune to their specific workflow.
• Firms with existing Anthropic enterprise relationships and Claude Pro/Team/Enterprise deployments already in place. Adding the Cowork plugin layer is incremental engineering, not net-new vendor procurement.
• Firms with deep internal precedent libraries and engineering capacity to build precedent-learning-equivalent features on top of the plugin. The customization payoff is real for firms at this profile.

The fit signal: if customization depth, $0 license cost, and integration with existing Anthropic infrastructure matter more than out-of-the-box deployment velocity, Cowork is the structural path. See the Anthropic Legal Ecosystem map for the broader build-your-own architecture.

Total cost of ownership: the comparison most firms get wrong

Spellbook does not publish a price: pricing is custom, based on team size, with a 7-day free trial (spellbook.com/pricing, checked October 2026). The naive comparison is wrong, in both directions, depending on firm profile.

For a 25-attorney mid-market firm without dedicated AI engineering staff:

• Spellbook license: a custom quote based on team size. Plus ~$5K implementation services. Plus minimal ongoing IT effort.
• Cowork plugin 12-month TCO: $0 license. Plus ~$80K-$150K for one mid-level AI engineer to deploy, configure, integrate with firm systems, and maintain. Plus Claude API costs at usage rates ($5/M input, $25/M output, per the pricing page), typically $5K-$15K annual at firm-scale usage. Plus integration development for Word add-in or alternative interface.

The Spellbook path is priced by written quote. The Cowork path costs ~$85K-$165K all-in. Spellbook may come out ahead on TCO at this firm profile once internal engineering time is counted; compare a written quote.

For a 200-attorney AmLaw 200 firm with existing AI engineering team and Anthropic enterprise relationship:

• Spellbook license: a custom per-seat quote that scales with headcount. Plus implementation services and ongoing per-seat scaling.
• Cowork plugin 12-month TCO: $0 license. Plus ~$50K-$100K incremental engineering capacity (the team exists, the Cowork plugin is one project among many). Plus Claude API costs at scale, possibly absorbed under existing enterprise commit. Plus integration with internal systems where engineering capacity is already deployed.

The Spellbook path is priced by written quote. The Cowork path costs ~$50K-$120K plus existing infrastructure. Compare a written Spellbook quote with the Cowork engineering and API cost at this firm profile.

The pivot point is firm size and internal AI capability, not the per-seat headline price. Below ~50 attorneys without internal AI engineering, Spellbook usually wins. Above ~150 attorneys with internal AI engineering, Cowork usually wins. The 50-150 attorney mid-tier is the most contested zone: the Spellbook pricing tier recommendations cover the per-firm-size fit math.

The second-order point: TCO comparison is incomplete without precedent-learning capability factored in. Spellbook Library is purpose-built. Cowork plugin doesn't have an equivalent native feature. Building precedent-learning on top of the Cowork plugin adds 6-12 months of engineering time to the TCO calculation. For firms where precedent learning is a core requirement, Spellbook's purpose-built capability often can justify the per-seat quote at firm sizes where Cowork would otherwise win on raw TCO.

Procurement and risk profile: what the build-vs-buy decision actually involves

Procurement velocity:

• Spellbook: Standard SaaS procurement. Typical timeline: 2-6 weeks from quote to deployment.
• Cowork plugin: Bypasses procurement entirely. IT can install without procurement approval. But integration, configuration, and rollout take 1-6 months depending on firm scale and customization needs.

Vendor risk:

• Spellbook: Vendor risk is real but mitigated by post-Series-B funding posture. Per the Series B funding analysis, the company has $80M+ cumulative funding and explicit growth targets: high vendor stability over a 2-year procurement window.
• Cowork plugin: No vendor risk per se (Anthropic could deprecate the plugin, but the open-source code remains available for self-hosting). Model-vendor risk is real: Anthropic API pricing or terms could change. Risk profile is structurally different.

Customization depth:

• Spellbook: Limited to playbook configuration and Library precedent training. Major customization requires vendor roadmap commitment.
• Cowork plugin: Unlimited customization. The trade-off is the customization is engineering work, not configuration.

Compliance and security:

• Spellbook: Vendor-managed compliance posture. SOC 2 and other certifications are vendor responsibility.
• Cowork plugin: Compliance posture is firm-managed. The plugin runs against the firm's chosen Claude deployment surface (Pro, Team, Enterprise, API, Bedrock, Vertex AI, Microsoft Foundry). Compliance follows the deployment surface choice.

Privilege defense (post-Heppner):

• Both paths require enterprise deployment surface for privilege risk mitigation. Per the Heppner SDNY ruling, consumer Claude generates non-privileged communications. Spellbook procurement at firm tier is enterprise-grade by default. Cowork plugin compliance depends on the underlying Claude deployment surface: Team/Enterprise or API minimum, not consumer Pro for privileged work.

The operational point: vendor risk is structurally lower for the Cowork plugin path because the source code is open, but model-vendor risk is structurally higher because pricing or terms could shift. Spellbook procurement carries vendor risk that's manageable post-Series-B but real over a 5-7 year horizon. Neither path is risk-free, and the risk profiles differ enough that comparison matters.

Decision framework: which path for which firm

Step 1: Do you have internal AI engineering capability or budget to hire it?

• Yes: Cowork plugin is on the table.
• No: Spellbook procurement is the structural path. Stop here.

Step 2: Is your contract review use case standard or non-standard?

• Standard (NDAs, MSAs, SOWs, employment agreements at volume): Spellbook is purpose-built and likely the right path even if you have AI capability.
• Non-standard (energy sector, life sciences IP, multi-jurisdictional regulatory): Cowork plugin's customization depth is the structural fit.

Step 3: Is precedent learning a core requirement?

• Yes: Spellbook Library is purpose-built. Building precedent-learning on top of Cowork is 6-12 months of engineering work.
• No: Cowork plugin handles core review and triage workflows without precedent learning.

Step 4: What's your firm size and TCO profile?

• Below 50 attorneys without dedicated AI engineering: Spellbook TCO usually wins.
• 50-150 attorneys: Most contested zone, depends on AI capability and customization needs.
• Above 150 attorneys with AI engineering team and Anthropic enterprise relationship: Cowork TCO usually wins.

Step 5: For Canadian firms specifically, CBA member status?

• Yes: Spellbook preferred-access pricing shifts the math in Spellbook's favor. Don't assume the discount is auto-applied: explicitly request it in the quote.
• No: Standard quote applies; Cowork plugin is a stronger relative comparison.

The synthesis: build-vs-buy decisions in legal AI follow firm capability profile more reliably than they follow firm size. A 25-attorney boutique with strong internal AI capability may rationally choose Cowork. A 150-attorney firm without that capability may rationally choose Spellbook. Firm size correlates with capability but doesn't determine it. The Spellbook 100M ARR business model analysis covers what the funding trajectory means for vendor procurement decisions over the next 24 months.

The bottom line. My take: Build-vs-buy in contract review tracks firm AI capability more than firm size. Spellbook is the right path for firms without internal AI engineering and with standard commercial contract use cases: Word add-in deployment, vendor-managed updates, predictable per-seat economics. Cowork plugin is the right path for firms with AI engineering capacity, non-standard contract types, or established Anthropic enterprise relationships: open-source customization with $0 license cost, but real engineering and maintenance overhead. The 50-150 attorney mid-tier is the most contested zone. Run the 7-day Spellbook free trial alongside a Cowork plugin pilot if you have the capability: direct comparison on your actual contract types reveals fit faster than headline price comparison.

Price sources, checked 22 October 2026: Claude (Anthropic), Claude API token rates. Each figure above is a list price from the vendor's own pricing page, or arithmetic on one. Plans, seat minimums and promotions change; confirm on the vendor's page before you buy.