5 Best Claude MCP Servers for Lawyers in 2026

Anthropic's May 2026 Claude for Legal launch shipped more than 20 MCP connectors for legal software, and by October 2026 the Model Context Protocol has become the standard way lawyers plug Claude into the systems where their work actually lives. But most of the coverage so far has centered on big-firm tooling: document management, contract repositories, and legal research databases. Criminal-defense practitioners have a different problem. Their 'system of record' is a discovery dump — bodycam video, interrogation footage, jail calls, 911 audio, police reports, scanned documents — and until recently there was no MCP server that let Claude query that evidence with citations you can verify. This list ranks five Claude MCP servers that genuinely help lawyers in 2026, with a bias toward defense-side workflows: evidence review, motion to suppress preparation, and grounded, auditable answers. Every entry uses only features confirmed by the vendor's own documentation, and every recommendation comes with the same caveat lawyers already know: an MCP server extends your reach, but it never replaces your judgment.

#1 Evidence Copilot — Solo $39/user/month (15 media-processing hours); Firm $99/user/month (60 hours, priority queue); PD Office $750/attorney/year (100 hours). Overage metered and hard-capped at 3x included hours.

Evidence Copilot is the only defense-side platform in this list with a Claude MCP server, and it earns the top slot because it connects Claude to the thing criminal lawyers actually need queried: the evidence. Upload a discovery dump — bodycam video, interrogation footage, jail calls, 911 audio, police report PDFs, photos, scanned documents — and the platform produces timestamped, diarized transcripts and OCR'd document text, then lets Claude query it all through five read-only tools: list_cases, list_evidence, search_evidence, read_window, and get_citation. Every answer Claude gives carries a citation into the exact second of video or page of the PDF, so you can verify anything in one click. The MCP server is read-only, org-scoped, uses revocable tokens, and is fully audited — the configuration defense lawyers should demand before connecting any AI to case materials. In practice this means you can sit in Claude and ask things like 'find every mention of the stop duration across the bodycam transcript and the police report' and get answers tied to sources, not model memory. Around the MCP server sits a full evidence-review workspace: auto-signals that flag candidate inconsistencies (report vs. bodycam, statement vs. call audio) before you ask, one-click case briefs, cross-exam drafts where every suggested question is tied to a quoted source, agent playbooks like Suppression Motion Prep and Miranda Timeline, frame-accurate video and audio redaction where originals are never touched, and multi-camera sync for bodycam plus dashcam plus interview-room footage. Every AI output is marked as AI-generated and pending human review, originals are immutable, and every action is audited. Published pricing with no sales wall: Solo at $39/user/month with 15 media-processing hours, Firm at $99/user/month with 60 hours and a priority queue, and a PD Office tier at $750/attorney/year with 100 hours. Overage is metered and hard-capped at 3x included hours, and every plan includes the full workspace — there are no feature tiers.

Best for: US criminal-defense attorneys, law clerks, defense investigators, and public-defender offices who want Claude to query actual case evidence — not just documents — with verifiable citations.

#2 Thomson Reuters CoCounsel Legal — CoCounsel is approximately $400/user/month; the MCP connector requires an active CoCounsel Legal subscription with MCP enabled.

Thomson Reuters shipped an MCP integration connecting Claude with CoCounsel Legal on May 12, 2026, and it is the most heavyweight research connector on this list. According to Thomson Reuters, CoCounsel Legal reasons across 1.9 billion Westlaw and Practical Law documents and 1.4 billion KeyCite validity signals, with a citation ledger designed to make every source traceable in one click. The company describes the next generation of CoCounsel Legal as rebuilt on Anthropic's Claude Agent SDK, positioning it as a research and drafting engine that lawyers can move work between and Claude. The connector itself is candidly scoped. Per Thomson Reuters' own connector guide, as of October 2026 it requires an active CoCounsel Legal subscription with MCP enabled, supports Claude Desktop, and exposes one capability: deep legal research. Its four tools start a research session targeting up to three U.S. jurisdictions, check progress, retrieve a cited report in markdown with Westlaw links, and ask follow-up questions against the finished research. Runs take several minutes for a concise report and twenty minutes or more for an expanded one. Two honest caveats for practitioners. First, none of the connector's four documented tools is a standalone KeyCite check, so confirming a case is still good law still requires a trip to the citator. Second, the ~$400/user/month cost of CoCounsel puts this well out of reach for most solo practitioners and public-defender offices — this is a connector for firms already invested in the Thomson Reuters ecosystem, not a starting point for defense-side evidence work.

Best for: Firms and legal departments already subscribed to Thomson Reuters research products that want Claude to run deep research against Westlaw and Practical Law.

#3 Harvey — Contact for pricing — requires a Harvey subscription with the account enrolled in Harvey's MCP feature.

Harvey publishes an MCP server, documented at developers.harvey.ai, that works with Claude (web and desktop), Google Gemini, and Microsoft 365 Copilot. Per its documentation, the server exposes a handful of tools: ask Harvey a general legal question, ask a question about documents in one of your Harvey Vault projects, run a question against a selected knowledge source, and list which Vaults and knowledge sources you can see. Authentication is per user over OAuth, and the server can only reach data the authenticated user already has permission to view in Harvey. Tool calls are stateless, with no conversation history retained between requests. In practice, the Harvey connector is a question-answering bridge rather than a control surface. Claude can pull a Harvey answer — with Harvey's citations — into a broader workflow it is running, such as an agent assembling a diligence memo that also reads a data room. But it cannot create Vault projects, upload documents into Harvey, run Harvey workflows, or edit anything on the Harvey side through the documented tools. The practical limitation is access. The connector presupposes a Harvey subscription with your account enrolled in Harvey's MCP feature, which places it well outside the reach of most solo practitioners and small firms. For firms that do have Harvey, it is a genuinely useful way to let Claude's agents consult Harvey's research layer as part of a larger workflow. For defense-side evidence work — bodycam, interrogation footage, discovery review — it does nothing, which is why it ranks below the evidence-focused options here.

Best for: Firms and legal departments already running Harvey that want Claude to consult Harvey's research and Vault content within larger AI workflows.

#4 Midpage — Contact for pricing — requires an active Midpage subscription.

Midpage runs a remote MCP server for U.S. case law that gives Claude access to a case-law database and an AI-powered citator — something general-purpose Claude, even with web search, cannot reliably replicate. According to Midpage's own announcement, the MCP gives Claude tools to run complex search queries directly over its comprehensive, up-to-date dataset of case law, enriched with proprietary metadata including the AI-powered citator so users can confirm cases are still good law. The company says the connector lets users answer legal research questions, generate first drafts of memos and briefs, connect facts from Claude projects to case law, review and cite-check briefs, and brainstorm or red-team legal strategies. The setup is straightforward: the MCP works in the Claude desktop application or in the browser, and requires an active Midpage subscription plus an active Claude Pro, Team, or Enterprise subscription. Notably, Midpage states the MCP is not part of any official partnership with Anthropic — it is a vendor-published server you connect yourself, which is normal in the MCP ecosystem but worth knowing when evaluating support expectations. For criminal-defense practitioners, Midpage covers the research half of the workflow — finding and validating case law for a suppression motion or a brief — but not the evidence half. There is no capability for querying bodycam transcripts, police reports, or recorded statements. Pair it with an evidence-focused platform rather than expecting it to handle discovery review.

Best for: Practitioners who want Claude connected to a case-law database with citator functionality for research, drafting, and cite-checking.

#5 CourtListener (Free Law Project) — Free.

CourtListener, run by the non-profit Free Law Project, offers open-access dockets and opinions through MCP at no cost, and it appears among the public-service connectors in Anthropic's May 2026 Claude for Legal launch. According to coverage of the launch, the CourtListener connector lets Claude search dockets and opinions — which makes it the lowest-risk way for a lawyer to see what an MCP-connected research workflow actually feels like before committing budget to a paid source. The case for CourtListener is simple: free access, public data, and low stakes. If you are a solo practitioner or a law clerk experimenting with your first Claude connector, you can connect CourtListener, ask Claude to pull opinions on a specific suppression issue in your circuit, and evaluate the workflow without any procurement conversation. The data is public case law, so there are no client-confidentiality concerns about what you are querying. The limitations are equally simple. CourtListener covers public dockets and opinions — not your case files, not your evidence, and not a citator with treatment signals the way paid research connectors do. You cannot use it to check whether a case has been negatively treated with the depth of a commercial citator, and it does nothing for the defense-side evidence problem of querying bodycam footage or police reports. Treat it as a free research on-ramp and a complement to, not a substitute for, the paid and evidence-focused connectors above.

Best for: Solo practitioners, law clerks, and public-defender offices piloting their first Claude connector with zero budget and zero confidentiality risk.

The right MCP server depends on which part of your practice you want Claude to reach. If your work is research-heavy and your firm already pays for Westlaw, the Thomson Reuters connector is the deepest option — just remember it does not replace a KeyCite run. Harvey suits firms already in its ecosystem, Midpage is a strong independent research layer with a citator, and CourtListener is the free on-ramp everyone should try first. But if your daily reality is a discovery dump — bodycam files with names like X13qrt9__x3455, interrogation footage, jail calls, and a police report that does not quite match the video — Evidence Copilot is the only MCP server on this list built for that job. Every answer Claude gives comes back with a citation into the exact second of video or page of the PDF, access is read-only and fully audited, and nothing is ever auto-applied: AI output is always marked for human review, and the platform never makes legal conclusions. Pricing is published, not hidden behind a sales call: Solo at $39/user/month, Firm at $99/user/month, and a PD Office tier at $750/attorney/year. Book a 20-minute walkthrough at evidencecopilot.com/book to see the workflow on your own kind of evidence, or start on the Solo plan at evidencecopilot.com/signup and connect the MCP server to Claude the same afternoon.

Frequently asked questions