Newcode

What it is

Newcode is a legal AI platform that a firm configures and runs as its own, rather than a finished assistant it rents. Newcode calls it an “AI harness”: a layer that connects the firm’s documents, systems, and models, then lets the firm build and govern its own agentic workflows on top. The firm picks the LLM, picks where it runs (Newcode’s cloud, the firm’s own Microsoft Azure tenant, or on-premises), and sets its own policies for what agents may do.

The product ships as three surfaces plus add-ins:

• Nova — the lawyer-facing assistant: referenced answers over firm data, Matrix Review for large document sets, DMS search, deep research, transcription, and a shared library of prompts and workflows per practice group. It also carries CCH Jurisdictional Compare, powered by Wolters Kluwer.
• Aurora — a drag-and-drop builder that chains LLMs, agents, and deterministic steps into workflows, with guardrails set centrally. A finished workflow publishes to Nova, to an API, or as an MCP server.
• Cowork — an agent that runs a task across systems end to end, for example pulling an agreement from Outlook and a playbook from OneDrive and reviewing one against the other. Newcode states Cowork deploys entirely in the customer’s own Azure cloud.
• Word, Outlook, and mobile apps, with named connectors for iManage, SharePoint, OneDrive, Teams, Excel, and PowerPoint, and 750+ MCP connections overall.

Maged Helmy, who holds a PhD in deep learning from the University of Oslo, founded the company in Norway in 2021. It raised a $6.5M seed round in March 2026 led by The LegalTech Fund and Alliance VC. On August 24, 2026 it announced a $13.5M Series A led by OnDean Forward, the fund of Relativity founder Andrew Sieja, with Relativity’s Rel Labs, The LegalTech Fund, and Antiportfolio Ventures participating. That makes $20M raised in 2026. Newcode reports 60+ law firms and government agencies as customers, ARR up more than 3x in 12 months, and a 60-person team in Dublin. Named customers include DLA Piper (a Nordic deployment announced June 2025), Reed Smith, Dickinson Wright, Robinson+Cole, and The Bar of Ireland.

Why legal ops teams look at it

The question Newcode answers is “buy Harvey, or build a firm-owned AI layer on top of iManage and our own precedents?” Three things drive teams to the build side:

• Model and hosting control. Firms with client data-residency terms, or an IT team that will only approve workloads inside the firm’s Azure tenant, can run Newcode there. The closed assistants run in the vendor’s cloud.
• Workflows as firm assets. An Aurora workflow for, say, a fund-formation checklist or an insurance-claims review belongs to the firm, can call any model, and can be exposed to other tools over MCP. That matters to firms that see their AI workflows as a product they sell to clients.
• Token economics. Because the firm chooses the model per step, it can route extraction to a cheap model and drafting to a frontier one. Newcode pitches this cost control directly.

Specific use case: a 150-lawyer firm with iManage builds one Aurora workflow that pulls a target’s contracts from the data room, extracts change-of-control and assignment clauses into a Matrix Review grid, checks them against the firm’s own M&A precedents, and publishes the workflow to Nova for every deal team.

Pricing reality

Newcode publishes no prices. Every deal is sales-led, and no credible third-party price point exists as of September 2026. The anchors in this segment are Legora at a reported $3,000 per user per year on a 10-seat minimum, and Harvey at a reported $12,000-14,400 per seat per year on a 20-25 seat minimum.

Budget three lines, not one: the Newcode license, the configuration work (Newcode embeds its own legal engineers with customers, as it did at DLA Piper), and model and Azure costs if you bring your own. Our planning estimate for a 25-50 lawyer first deployment is $40,000-120,000 in year one across all three lines. That is an estimate, not a quote. Ask Newcode to price the license per user and the legal-engineering work as a separate, capped line item.

Best for

Innovation and knowledge-management leads, and legal ops directors, at firms of roughly 50-1,000 lawyers that already run iManage and Microsoft 365, have at least one person who can own workflow design, and need AI to run inside their own Azure tenant or on a model they choose. Legaltech Hub lists 20+ lawyer firms and corporate legal departments as its target.

Alternatives and when to pick them

• Harvey — the market-share leader. Pick it when you want a finished product with the deepest AmLaw adoption and your IT team accepts vendor-hosted processing.
• Legora — the fastest-growing rival, with Word-native review and Tabular Review. Pick it for a 10-200 seat team that wants to buy, not build, at roughly a fifth of Harvey’s per-seat price.
• Thomson Reuters CoCounsel — pick it when research grounded in Westlaw content matters more than custom workflows.
• Wordsmith — pick it for an in-house team that wants self-serve legal agents without a configuration project.
• Build on n8n plus a model API — pick it only if you have engineers, not lawyers, owning the build, and accept that you also build permissions, audit logs, and a lawyer-facing interface yourself.

If none fit, run Claude or Microsoft Copilot on an enterprise plan with a written use policy and a shared prompt library for 90 days, and measure which tasks lawyers repeat before you buy a platform.

Watch-outs

• Configurability is a workload. One independent review site scores ease of use at 4.6/10, its lowest dimension, and the value depends on someone designing workflows. Guard: name an internal workflow owner before signing, and contract for 3-5 production workflows delivered by Newcode’s legal engineers in the first 90 days, with acceptance criteria.
• Your model bill is now your problem. Model choice means model cost and model risk sit with the firm. Guard: set a token budget per workflow, have the cost per matter of each pilot workflow reported monthly, and document which model each workflow uses for your client AI disclosures.
• No certifications on the public record. We found no SOC 2 or ISO 27001 statement on Newcode’s public pages; its trust center is gated. Guard: request the SOC 2 Type 2 report, the subprocessor list, and the model providers’ data-retention terms before any client data is connected.
• A small vendor scaling fast. $20M raised is a fraction of Harvey’s or Legora’s war chest. Guard: favor the Azure-tenant or on-premises deployment so data and workflow definitions stay with you, and get a contractual right to export every Aurora workflow in a documented format at exit.

Background: MCP server explained and retrieval-augmented generation.