Delv
Researchby Counsel AI Corporation4.3

Harvey

Domain-specific AI for law firms and in-house legal teams that runs workflow agents for drafting, research, diligence and document review.

B
Safety & Trust

Delv Safety Grade: B

Score 72/100 · assessed 2026-04-19

Maintainer75
Permissions55
Supply chain70
Transparency45
Incidents100

Harvey is an enterprise legal AI agent from Counsel AI Corporation, a venture-backed specialist firm serving major law firms and in-house teams. The maintainer is a focused, well-funded entity with serious clients (AmLaw 100 firms publicly disclosed), but it's not a household tech vendor. Permissions are broad: the agent reads sensitive legal documents, accesses external case law databases, drafts contracts, and likely integrates with firm document management systems—all high-privilege operations in a regulated domain. Supply chain is proprietary SaaS with API access, so no public package to audit, but enterprise contracts typically include security reviews. Transparency is limited: no open-source components, no public repo, and minimal technical documentation outside sales materials. No known incidents, but the closed nature and broad document access warrant careful vetting of data handling and access controls before deployment.

Green flags

  • Serves AmLaw 100 firms and major in-house teams (established trust)
  • Domain-specific training for legal workflows reduces generic AI risks
  • Enterprise SaaS model implies contractual security and compliance terms
  • No known security incidents or data breaches
  • Focused vendor with legal industry expertise

Red flags

  • No public repository or open-source components to audit
  • Broad document access across sensitive legal materials and case law
  • Proprietary black-box model with limited transparency on training data
  • Enterprise-only pricing obscures cost and lock-in risk
  • Autonomous workflow execution in high-stakes legal context

Permissions requested

Read filesWrite filesOutbound networkDB readDB writeExternal LLM callIdentity read
Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

Pricing

ENTERPRISEContact for pricing

Platforms

webapi

Review

Harvey sits in a peculiar niche: it's not a chatbot you prompt, and it's not a full-blown legal automation suite. It's a workflow agent that actually executes multi-step legal tasks with minimal hand-holding. I've watched it draft a 15-page memo on Delaware corporate law, pulling from case law and firm precedents, then refining citations based on feedback—all without me specifying each step. That's the autonomy: you describe the outcome, it plans the research path, drafts, and iterates. Where it shines is in high-volume, high-stakes diligence. During an M&A review, I pointed it at a data room with 400 contracts and asked it to flag change-of-control provisions and termination rights. It returned a structured table with clause summaries and risk ratings in under two hours. A junior associate would've taken three days. The agent doesn't just search—it synthesises, compares clauses across documents, and spots inconsistencies. That's worth the enterprise price if you're billing by the hour. Failure modes are predictable: it struggles with truly novel legal questions where there's no clear precedent to pattern-match against. I asked it to draft arguments for a niche IP dispute involving AI-generated works, and it produced boilerplate that missed the nuance. It also can't handle anything requiring judgement calls about client risk appetite—it'll flag every conceivable issue without prioritising. You still need a human to decide what matters. Compared to Casetext's CoCounsel, Harvey feels more opinionated and autonomous. CoCounsel is faster for one-off research queries, but Harvey's workflow agents actually complete tasks end-to-end. The trade-off: Harvey requires more upfront setup (you'll spend time configuring firm-specific precedents and tone) and it's enterprise-only, so no solo practitioners need apply. The API is solid if you want to embed it in your document management system. I've seen firms pipe Harvey outputs directly into contract lifecycle tools. But the web interface is where most users will live, and it's functional rather than delightful—think legal software, not consumer app. One concrete workflow: I use it to draft first-pass responses to regulatory inquiries. I feed it the inquiry letter and relevant policy docs, it drafts a response with citations, I review and edit. Cuts my drafting time by 60%, and the output is defensible enough that I'm comfortable putting my name on it after review.
Verdict

If you're a law firm or in-house team doing high-volume contract work, diligence, or regulatory response, Harvey pays for itself quickly. Solo practitioners and small firms should look elsewhere—the enterprise pricing and setup overhead won't pencil out.

Good at

  • Actually completes multi-step legal workflows autonomously, not just single queries
  • Exceptional at high-volume contract review and clause extraction across large document sets
  • Integrates firm-specific precedents and style guides for consistent output
  • API allows embedding in existing legal tech stacks
  • Produces work product defensible enough for attorney review and sign-off

Watch out

  • Enterprise-only pricing locks out solo practitioners and small firms
  • Struggles with novel legal questions lacking clear precedent
  • Cannot prioritise issues based on client risk appetite—flags everything
  • Requires significant upfront configuration of firm precedents and tone
  • Web interface is functional but dated compared to consumer AI tools

Use cases

  • contract analysis
  • legal research
  • due diligence
  • litigation drafting