Spellbook

Spellbook is an AI contract assistant from Canadian legal-tech company Rally, and it was one of the earliest tools to build on OpenAI’s GPT-4 for legal drafting. Its defining choice is where it lives: rather than another web platform to log into, Spellbook installs as a Microsoft Word add-in, appearing in the sidebar of the tool lawyers already use to write and review agreements.

Inside a document, it reviews contracts and flags unusual or risky language, points out clauses that are missing relative to a standard agreement, drafts new provisions from a plain-English request, and generates redlines. Newer “Associate” agentic features can run a fuller review pass across a document. It competes with tools like Luminance, LinkSquares, and Harvey, but leans toward small and midsize firms and in-house teams who want assistance without adopting a heavyweight contract-lifecycle platform.

Key Features

• Microsoft Word add-in — no separate interface to learn
• Contract review with risk and unusual-term flagging
• Missing-clause detection against standard agreements
• Clause and full-provision drafting from natural language
• Redline generation, including against custom playbooks
• Benchmarking and negotiation suggestions on common terms

Pricing

• Subscription (per seat): Typically around $99/user/month for the core plan, usually billed annually
• Enterprise: Custom pricing with volume discounts, SSO, and security review

Confirm current tiers directly, as legal-AI pricing changes often.

Best For

Transactional lawyers, in-house counsel, and small-to-midsize firms that review and draft contracts regularly and want AI help without leaving Microsoft Word.

Limitations

The per-seat price is a real commitment for solo lawyers or very small teams. It’s tied to Word, so Google Docs users are out of luck. And like any generative tool, its suggestions require careful human review — it can misread context or propose language that doesn’t fit the deal. For large enterprises with complex contract-lifecycle needs, a dedicated CLM platform may be a better fit.