What is MCP and how does it work?

Model Context Protocol is an open standard for connecting AI tools to external systems. When you connect Lexnus to your AI model via MCP, the AI model can query your clause library, evaluate clauses, and run compliance checks as part of its responses, without you switching applications.

Does Lexnus work with any AI model?

Yes. The MCP standard is open and provider-agnostic. Any AI tool that supports MCP can connect to Lexnus's policy engine. Lexnus is designed to work with any AI layer, not exclusively with Anthropic's products.

What is the difference between an AI model generating a clause and retrieving one from Lexnus?

When the AI model generates a clause from scratch, the output is new language that has never been reviewed by your organisation. When the AI model retrieves a clause from Lexnus, it returns language that has been reviewed, approved, and published by a qualified lawyer at your organisation. The first is fast but ungoverned. The second is fast and governed.

Does using Lexnus via MCP replace the Lexnus web application?

No. The MCP integration is a second access point to Lexnus's policy engine, one that works inside AI tools like your AI assistant in Word. The web application is where you build and manage playbooks, review clause submissions, and monitor compliance health. Both work together.

Features

MCP Integration: AI Tools and Lexnus

Lexnus exposes its policy engine and clause library as an MCP (Model Context Protocol) server. Your AI assistant in Word, any MCP-compatible AI interface, and any MCP-compatible agent can call Lexnus directly to evaluate clauses, retrieve approved language, and check compliance against your playbook in real time. General-purpose AI becomes safe for professional legal use not by restricting what it can do, but by giving it a governed policy layer to work against.

What MCP is and why it matters

Model Context Protocol (MCP) is an open standard for connecting AI tools to external systems. When a lawyer is working in Word with their AI assistant, the AI model can call Lexnus's MCP server to look up approved clauses, evaluate a specific provision, or run a compliance check, without the lawyer leaving their workflow or switching applications.

The result: AI drafting grounded in your organisation's approved language, not generated from scratch.

How the Lexnus MCP server works

↗

The architecture

The AI model operates as the AI layer. Lexnus operates as the policy layer. They are separate and complementary:

The AI model reads, interprets, drafts, and explains legal text.
Lexnus stores your approved clauses, enforces your rules, and determines whether any given clause or contract complies with your policy.

When the AI model is connected to Lexnus via MCP, it grounds its outputs in your organisation's standards. It asks Lexnus what your approved liability clause says. It checks whether the termination notice in a counterparty's draft meets your minimum. It retrieves the right jurisdiction-specific variant of a clause for the deal at hand. The AI model provides the language intelligence. Lexnus provides the policy authority.

⚙

Available tools

Lexnus's MCP server exposes 27 tools. The main ones, grouped by what the lawyer is actually doing:

Drafting — lexnus_resolve_playbook maps a plain-language request ("an NDA for a Swedish supplier") to a published playbook. lexnus_start_assembly and lexnus_answer_session_field collect the deal facts one at a time. lexnus_generate_document renders the contract from approved clauses.
Analysis — lexnus_ingest_contract takes a contract shared in chat. lexnus_analyse_document runs it against the playbook. lexnus_get_analysis_findings returns a verdict per rule, with the excerpt that triggered it and whether it blocks signature.
Approved language — lexnus_search_clause_library searches your published clauses. lexnus_get_playbook_clauses returns the clause structure of a contract type. lexnus_get_playbook_rules returns the rules that apply to it.
Counterparty revisions — lexnus_upload_version adds a redline as a new version, parses tracked changes, and reports edits made with revision tracking switched off.
Signature — lexnus_send_for_signature sends a reviewed contract through your signing provider. lexnus_get_signing_status tracks who has signed.

Full tool reference, with the scope each one requires: Lexnus MCP server documentation.

🔒

Authentication and audit

MCP connections authenticate via OAuth, or via a personal access token issued in Lexnus. Both carry scopes: untick write access when you approve an OAuth connection, or issue a token scoped to read, and every write tool rejects it. Every tool call, read or write, is recorded to Lexnus's append-only audit trail under the tool's name, stamped as having arrived over MCP, so you can separate AI activity from web activity. Writes also record the record touched and its before-and-after state. Full detail, including what the audit log does and does not cover: Lexnus MCP server documentation.

The tools that reach a counterparty by email — sending for signature, cancelling, reminding — are gated by Lexnus itself. The first call stops before the signing provider. A current client puts an explicit confirmation to you, and the model cannot approve that for you. An older client gets a signed confirmation token that covers exactly the contract and recipients it was shown. How the gate works.

What the MCP integration exposes

Clause grounding — Instead of generating novel legal language, the AI model retrieves your organisation's approved clauses from the Lexnus library. AI drafting stays within your governed standards.
Real-time compliance checking — During drafting or review, the AI model can call Lexnus to check whether a specific clause or provision passes your playbook rules before it goes anywhere.
Audit trail on every call — Every tool call an AI client makes, and every change it causes, is recorded to the append-only audit trail, stamped as having arrived over MCP. Contract versions added over MCP are separately flagged as agent-initiated, so the version history shows which changes came from an AI and which from a person.
Account and workspace scoping — MCP tools are strictly scoped. Cross-account data access is architecturally prevented.
Provider agnostic — The MCP integration is not exclusive to any single AI provider. Any MCP-compatible AI tool can connect to Lexnus. The policy layer works regardless of which AI you use.

Who connects AI tools to Lexnus

Lawyers using AI tools in Word get their AI drafting assistant grounded in the organisation's approved language. When they ask the AI model to draft or fix a clause, the AI model calls Lexnus to retrieve the right approved version rather than generating something new.

Legal Ops teams building automated contract workflows can use Lexnus's MCP server as the policy enforcement step in any agentic pipeline. An agent that generates draft contracts calls Lexnus before the output reaches anyone.

General Counsel gain oversight of what AI tools are doing with their organisation's contracts. Every call an AI tool makes, and every change it makes to a Lexnus-governed contract, is in the audit log, marked as having come over MCP.