Newcode AI-native law firm stack [Explained]

TL;DR

• A Newcode AI-native law firm stack is a firm-owned configurable AI harness: it connects the systems you already run, retrieves across matter files, lets lawyers build their own agents, and keeps permissions, model choice, and deployment under the firm.

• As of 24 August 2026, Norway-founded Newcode announced a Series A that Artificial Lawyer put at $13.5m, with Relativity's Rel Labs in the round, taking 2026 funding to $20m after a March seed.

• The mechanism is a four-step loop — Connect, Contextualize, Create, Control — on 700+ to 750+ Model Context Protocol (MCP) connections, plus two named products: Nova for grounded work product and Aurora for agentic workflow building.

• A 2-truck HVAC shop, a 10-person agency, or a solo clinic should not copy the logo; copy the pattern. Keep your system of record. Own the intelligence layer. Do not rent a black-box "AI lawyer" that you cannot inspect.

Key Takeaways

• The round is a capital event around a configurable legal AI harness, not a merger into RelativityOne and not a replacement for practice-management software such as Clio, Smokeball, or MyCase.

• MCP is the constraint that broke: instead of a custom connector for every DMS, mail system, and matter database, firms can plug tools into one open protocol that Anthropic open-sourced on 25 November 2024.

• Rel Labs joining the round matters because Relativity already ships its own RelativityOne MCP connector and counts 200,000+ product users; the bet is on an open bus, not a single chatbot.

• Small practices already live the failure mode Newcode is selling against: intake in one tool, files in another, AI in a third tab with no audit trail. The stack is the decision to stop treating those as separate products.

• Honest limits remain: Newcode has not published ARR dollars, token prices, or an independent accuracy study; EU AI Act transparency rules applied from 2 August 2026; the FTC has already fined an "AI lawyer" claim that could not stand up.

What a Newcode AI-native law firm stack is

A Newcode AI-native law firm stack is a firm-owned, configurable AI harness that sits on a practice's existing case data, mail, and document systems so lawyers can build their own agents and workflows instead of renting a vendor's fixed AI menu.

That sentence is the whole product idea. Everything else is funding, connectors, and governance. If you run a 2-truck HVAC company, a 10-person marketing agency, or a solo clinic, you already feel the same bind in operational terms: the job ticket, the client brief, or the chart lives in one system, the inbox lives in another, and the "AI helper" you tried last quarter cannot see either without a paste. You do not need a Norway legal-tech Series A to recognize the cost. You need a layer that talks to the tools you already pay for, keeps permissions at the job/matter/patient level, and lets you swap models without rebuilding intake.

The legal version of that layer is what Newcode is selling. On its homepage the company calls itself "The Configurable AI Harness for the Legal Industry" and lists named customers that include Kirkland & Ellis, DLA Piper, Reed Smith, Dickinson Wright, Wolters Kluwer, Robinson + Cole, The Bar of Ireland, and several Norwegian firms. The product overview frames the same problem in operator language: email sprawl, context loss, inconsistent outputs, governance friction, fragmented workflows, decentralized data. The promised fix is not another chatbot. It is centralized intelligence so context follows the matter rather than the tool.

That is why this term belongs on a small-business desk, not only an AmLaw knowledge-management committee. A clinic that cannot let a model read the wrong chart has the same wall-crossing problem as a firm that cannot let a junior see the wrong matter. An agency that must keep one client's creative files out of another client's deck has the same permission problem as a conflicts wall. The Newcode AI-native law firm stack is the legal-market name for "own the harness, keep the system of record." Teams that already route documents through US Tech Automations can treat that harness as a model-and-connector swap rather than a practice-management rebuild.

What happened on 24 August 2026

On 24 August 2026, according to Artificial Lawyer, Newcode raised $13.5 million Series A in August 2026, with Relativity's investment arm Rel Labs joining the round.

The same report named OnDean Forward, Antiportfolio Ventures, and TLTF (The LegalTech Fund) as joining, and put 2026 capital so far at $20m after a seed of around $6.5m in March. US expansion is listed as a use of proceeds. Founder and CEO Maged Helmy's published comment is that the legal market needs control and flexibility to become AI-native, and that the mission is to help firms build AI capabilities that are truly their own.

Legal IT Insider covered the same day and was more cautious on the Series A headline number. It stated that Newcode had not publicly stated the value of this Series A, while still reporting the March seed and the 2026 total. According to Legal IT Insider, the round brings 2026 funding to $20 million raised in 2026. According to Legal IT Insider, Newcode raised $6.5m in seed funding in March. The outlet said the Series A was led by OnDean Forward with significant participation from The LegalTech Fund (also a seed investor), alongside Antiportfolio Ventures, Rel Labs, and other existing investors.

Those two write-ups agree on date, lead, seed, total, and Rel Labs. They disagree on how explicitly the Series A check was announced. This hub treats $13.5m as Artificial Lawyer's reported Series A figure and $20m / $6.5m as the overlapping 2026 capital picture both outlets published. The arithmetic that ties those three numbers together is in the USTA analysis section below; it is not a Newcode press-kit number.

Legal IT Insider also published operating claims the Series A story hangs on: more than 60 law firms and government agencies across Europe and the United States as of that announcement, annual recurring revenue more than tripled in the last 12 months, new customers across Europe, the AmLaw 200, a growing base of US deployments, and The Bar of Ireland, which the article said provides support to more than 2,150 barristers. As of this writing, Newcode's own homepage instead says "100+ law firms and enterprises." Both figures are vendor-adjacent; neither is an audited customer count. Use them as a range, not a census.

Andrew Sieja — Relativity founder and principal partner at OnDean Forward — is quoted in both outlets saying the tech gives law firms a path to own their AI stack and accelerate AI across the practice and business of law. Zach Posner, managing partner at The LegalTech Fund, is quoted in Legal IT Insider on a single point he says firm leaders repeat: they want platforms that adapt to how they operate, not platforms that force them to adapt. TLTF's own site states 80+ portfolio companies, 1,000+ intros per year, 120+ fund advisors, and 25+ years of general-partner experience.

Sources: Artificial Lawyer; Legal IT Insider.

The mechanism: Connect, Contextualize, Create, Control

Artificial Lawyer printed Newcode's four support areas in the Series A piece. The vendor homepage uses the same four verbs. Read them as an operations loop, not a slogan.

Connect. According to Artificial Lawyer, Newcode offers 700+ MCP connections and a comprehensive API so firms can link to any data source. Newcode's own site, fetched for this hub, now says "750+ MCPs", naming Outlook, iManage, Teams, and OneDrive as the daily-work examples. iManage is not a niche add-on: the DMS vendor states 4,000+ global customers, 1 million users worldwide, 42% of the Fortune 100, and 81% of the AmLaw 200. A harness that cannot see the DMS is a demo. A harness that can search it under existing permissions is a stack.

Contextualize. Newcode says it has built a proprietary agentic retrieval system that plans, searches, evaluates, and iterates until the context is right, guided by firm-defined taxonomy, labels, and walls. That is the difference between "chat with a PDF" and "retrieve across the matter with the wall still on." The overview page calls this connected intelligence: every source, model, and workflow reporting into one system so context follows the matter.

Create. Users build work product across those sources or reach Newcode through a frontier model. The named create surface is Nova: retrieval and analysis with referenced answers grounded in firm data. Nova's published modules include matrix review of large document sets, a reusable knowledge-base assistant, CCH Jurisdictional Compare powered by Wolters Kluwer, legal research, deep research, DMS search, transcription, a prompt library, personalization, and a memory feature that Newcode says does not store private or confidential data and never uses user information to train or fine-tune models. Those last two clauses are vendor statements, not an audit.

Control. Firms get granular permissions, cloud or local deployment, and the ability to use whatever LLM and interface works for them. The builder surface is Aurora: drag-and-drop agentic workflows, deploy-from-one-place, firm policies and guardrails, and an option to turn a workflow into an application for internal and external stakeholders. Control is the part small shops skip, then regret, when a model answers from the wrong client's folder.

Helmy's Artificial Lawyer interview adds one more operational claim that is easy to over-read: token costs can spiral on complex workflows and high-volume document work; Newcode says it provides transparency, cost controls, routing logic, and optimisation, and that the result is predictable cost behaviour and in many cases significant reductions in spend. No dollar figure or percentage is attached. Do not treat that sentence as a savings calculator.

Sources: Artificial Lawyer (700+); Newcode (750+, 100+); Legal IT Insider (60+); Rel Labs (200,000+); Relativity (25 years).

Why the constraint broke: MCP as the bus

The reason a "configurable legal AI harness" is fundable in 2026, rather than another closed legal chatbot, is that the connector problem got a standard.

Anthropic open-sourced the Model Context Protocol on 25 November 2024 as a universal way to connect AI systems with data sources, replacing fragmented integrations. The MCP documentation describes it as an open-source standard for connecting AI applications to external systems — local files, databases, tools, workflows — and uses the USB-C analogy. The GitHub organization hosts the specification and official SDKs; as of this fetch the servers repository showed 90,011 stars and the org listed 50.3k followers. Those star counts move. They are evidence of ecosystem weight, not of legal-grade security.

Relativity did not sit the protocol out. Its MCP product page describes a RelativityOne connector that lets assistants such as Claude, Gemini, and Copilot perform setup and operations work — workspaces, access review, usage reporting — while keeping RelativityOne as the system of record. Relativity is explicit that MCP is for orchestrating the work around a matter and Relativity aiR is for the analysis inside it. Relativity's aiR page cites customer results it attributes to named firms: Microsoft expects to save an average of 20%–40% on document review with generative AI; Teneo reviewed 1 million documents in 18 days, yielding 70% cost savings; KordaMentha cut costs by 85% while saving 25 days of review time. Those are Relativity-published case claims, not Newcode metrics, and they describe a different product. They matter here because Rel Labs is now on both sides of the bus: a system-of-record vendor shipping MCP, and a check into a harness vendor that leads with 750+ MCP connections.

Rel Labs itself is Relativity's legal-tech innovation and investment program. It advertises hands-on support with The LegalTech Fund, access to Relativity's platform, built-in reach through 200,000+ professionals using Relativity products worldwide, and potential capital. Relativity CEO Phil Saunders is quoted there on community and scale. Rel Labs also states that it encompasses the developer program and the strategic investment motion, including the partnership with The LegalTech Fund. A Rel Labs check in Newcode's Series A is therefore not a random late-stage logo. It is the e-discovery incumbent putting money on an open connector layer that its own platform already speaks.

For a small practice the lesson is narrower than "buy Newcode." The lesson is that the integration tax that used to make firm-owned AI a Big Law project is the thing MCP is designed to cut. If your matter files already live in a DMS, your mail in Microsoft 365, and your time in a practice-management system, the stack question is which harness talks to those systems under your permissions — not which chatbot has the flashiest demo.

Who shipped it and who is already in the adjacent stack

Newcode is Norway-founded. Helmy is the named CEO. The products on the public site are Nova and Aurora, not a single "robot lawyer" SKU. That distinction is not cosmetic. On 25 September 2024 the FTC announced Operation AI Comply, five law-enforcement actions against operations using AI hype or selling AI that can be used deceptively. One of those actions was against DoNotPay for claiming to be "the world's first robot lawyer" and offering services the complaint said could not substitute for a human lawyer; the proposed order included a $193,000 payment and a ban on professional-substitution claims without evidence. A Newcode AI-native law firm stack that markets control, permissions, and human-configured workflows is playing a different game than an "AI lawyer" button. Keep it that way in your own copy.

The adjacent stack most small US firms actually run is still practice management plus documents, not an AmLaw knowledge platform. Clio states that 400,000+ legal professionals trust its system, with 18+ years in the market, 100+ bar approvals including all 50 states, 300+ app integrations, and 1 billion+ legal docs from 100+ countries. Smokeball states it is trusted by 34,000+ legal professionals worldwide, and prints two workload figures on its homepage: 40% of legal professionals lose over a third of their time to non-billable work, and 30% of fee earners bill less than half of their working day. MyCase states 19,000 law firms choose MyCase and that it is partnered with over 130+ bar associations. If you are still choosing among those systems, start with the live comparisons for Clio alternatives for solo lawyers, Smokeball vs Clio Manage for transactional practices, and MyCase vs Clio Manage for family law. A harness that cannot read the matter system you already picked is not a stack. It is another tab.

Clio's research is the cleanest public window on why firms even want a harness. The Legal Trends landing page says growing firms are growing revenue 4x faster than headcount, meeting the 50%+ of clients who now turn to AI first, and reducing mental strain by up to 25% with technology. The 2025 Legal Trends Report documents the method: aggregated data from tens of thousands of US legal professionals, a survey of 1,702 US legal professionals, a survey of 1,000 US adults, 24 interviews, diary studies, and a neurological study of 63 legal professionals. According to Clio's 2025 Legal Trends Report, growing firms increased lawyer headcounts by 25% while revenues increased four times as much. The same report says growing firms in its cohort doubled revenues over four years with only a 50% increase in clients and matters, and that growing firms use AI in Clio twice as much as stable and shrinking firms. Those are Clio's cohort findings, not Newcode's.

Sources: Clio; Smokeball; MyCase; iManage.

USTA analysis: how the 2026 checks stack

USTA analysis (derived only from figures cited above). Inputs: Series A $13.5m as reported by Artificial Lawyer; seed $6.5m as reported by Legal IT Insider and described as around $6.5m by Artificial Lawyer; 2026 total $20m as reported by both. Arithmetic: $13.5m ÷ $6.5m = 2.08, so the August check is 2.08 times the March seed check. $13.5m ÷ $20m = 0.675, so the Series A is 67.5% of published 2026 capital. March 2026 to 24 August 2026 is roughly five months; the exact seed closing day was not published in the two announcement pieces, so this hub does not claim a day count. What the arithmetic does support: Newcode concentrated most of its 2026 capital in the Rel Labs-inclusive round, not in the seed, and it did so on a short clock after the first check.

That is a pace signal, not a valuation. Neither outlet published a post-money number, an ARR dollar figure, or a price per seat. Legal IT Insider's "ARR has more than tripled" claim has no baseline, so this analysis does not convert it into a revenue multiple. The 60+ (24 August, Legal IT Insider) versus 100+ (vendor homepage at fetch) customer-language gap is also not a growth rate; the two phrases are not identical ("law firms and government agencies" vs "law firms and enterprises"). Do not annualize them.

What a 2.08× follow-on check in about five months does tell a small-firm buyer is simpler: investors paid to scale a harness story, with US expansion named as a goal, before the vendor had published the usual SaaS scoreboard. Budget for implementation labor and for connector review. Do not budget as if a public ARR multiple were in the announcement.

What it changes for small practices

The change is not "fire the practice-management vendor." The change is where intelligence is allowed to live.

If your system of record is Clio, Smokeball, MyCase, or a folder-plus-DMS setup, a Newcode-style stack says the AI layer should be configurable on top of that record: connect the sources, retrieve with walls, create work product with citations, control models and permissions. That is the opposite of pasting a confidential intake into a consumer chatbot, and it is the opposite of waiting for one vendor's quarterly AI feature drop.

A 10-person agency that files statements of work the way a clinic files charts can reuse the same extraction pattern US Tech Automations already runs on invoices and intake forms. The HVAC analog is even plainer: the job packet (photos, invoices, equipment notes) is the matter file. The dispatch board is the practice-management system. The "AI" that cannot see either is a toy. The harness that can read both under role permissions is the stack.

Relativity remaining the e-discovery system of record, with its own MCP connector, is the Big Law version of the same split: do not move evidence out of the place that already has the audit trail. Relativity's MCP page repeats that RelativityOne stays the system of record and that every MCP action is captured in Relativity Audit and attributed to the authenticated user. Small firms rarely have RelativityOne. They do have a matter system, an inbox, and a file store. The Control pillar — local or cloud, model choice, granular permissions — is the part to copy even if the vendor you pick is not Newcode.

Governance is not optional color. The NIST AI Risk Management Framework 1.0 (January 2023, NIST AI 100-1) is voluntary and organizes work into GOVERN, MAP, MEASURE, and MANAGE. NIST's AI RMF page notes the 26 January 2023 release, the 26 July 2024 generative-AI profile (NIST-AI-600-1), and an 7 April 2026 concept note for a critical-infrastructure profile. ISO/IEC 42001:2023 is the AI management-system standard, published December 2023, 51 pages. The UK ICO guidance on AI and data protection, updated 15 March 2023, is under review after the Data (Use and Access) Act and still centers accountability, transparency, lawfulness, accuracy, and fairness, including Article 22 of the UK GDPR on solely automated decisions. Norway-founded or not, a vendor selling into the EU now sits under the EU AI Act: the Commission says the Act entered into force on 1 August 2024, became applicable on 2 August 2026, defines four risk levels, lists nine prohibited practices, and puts AI used in the administration of justice on the high-risk list, with high-risk obligations described as starting 2 December 2027 after the Omnibus timeline change. Transparency rules are described as coming into effect in August 2026. Those dates are why "cloud or local" and "granular permissions" are product features rather than slides.

Honest limits

Newcode has not, in the two announcement pieces, published seat price, token price, ARR dollars, gross margin, or an independent hallucination study. Helmy's token-cost comments are qualitative. The 700+ vs 750+ MCP gap is a vendor-count moving target; "MCP connection" is not a regulated unit. Named logos on a homepage are not the same as a referenceable case study with hours saved.

A configurable harness can still leak if permissions are wrong. Agentic retrieval that "iterates until the context is right" can iterate its way across a wall if the wall is mis-tagged. Local deployment reduces one class of vendor-cloud risk and increases a class of endpoint-and-backup risk. Model choice is only a control if someone at the firm owns routing rules; otherwise it is a menu.

Practice-management AI is not the same product. Clio, Smokeball (Archie), and MyCase (8am IQ) all ship AI inside the matter system. Newcode is pitching the layer above and across systems. Buying both without a system-of-record decision gives you two agents arguing over the same file.

DoNotPay's FTC case is the outer bound: claiming an AI service substitutes for a lawyer, without testing that it does, is a chargeable claim in the United States. Newcode's public language is "build AI capabilities that are truly their own," which is a platform claim, not a license to practice. Keep a human on every client-facing output. Relativity's own aiR FAQ states that fully autonomous agents without human oversight do not belong in legal tech. That sentence is a useful default even if you never buy Relativity.

Signal vs Speculation

Fact (sourced): On 24 August 2026 Newcode announced a Series A led by OnDean Forward with Rel Labs, The LegalTech Fund, and Antiportfolio Ventures in the round; Artificial Lawyer reported $13.5m; both outlets reported $20m total 2026 funding and a $6.5m March seed; Newcode publicly describes a Connect–Contextualize–Create–Control harness, Nova, Aurora, and 750+ MCP connections; Relativity ships RelativityOne MCP and Rel Labs; MCP is an open protocol Anthropic released on 25 November 2024; EU AI Act applicability is dated 2 August 2026 on the Commission page.

Our read: If MCP stays the default way assistants talk to DMS, mail, and matter systems, then over the next 12–36 months small and mid-size professional firms — law first, then adjacent document-heavy shops — will be forced to pick a harness strategy even if they never hear the word Newcode. The likely split is: system of record stays (Clio/Smokeball/MyCase/iManage or a vertical equivalent); the intelligence layer becomes configurable; point-solution "AI buttons" that cannot honor walls get demoted to drafts. Rel Labs in this round makes it more likely that RelativityOne remains the e-discovery record while a separate harness sits on top, not that Newcode gets folded into aiR.

Our read (SMB path): A 10-lawyer firm or a 10-person agency will not stand up 750 connectors. It will stand up five: mail, files, matter/job system, billing, and one research or knowledge source. The Newcode story still applies because the design choice is the same. Firms that treat AI as a rented feature inside one vendor will keep waiting on that vendor's roadmap. Firms that treat AI as a governed layer they own will look like the Clio "growing" cohort: more revenue per head, not more tools per head.

Our read (risk path): If EU high-risk duties for justice-administration tools land as written, and if US state bars keep treating unsupervised generative output as a competence and confidentiality problem, then vendors that cannot show matter-level permissions, audit logs, and human-in-the-loop will stall in procurement. That favors the Control pillar over the Create pillar in any stack that wants to survive a client audit.

Glossary

• Newcode AI-native law firm stack: A firm-owned configurable AI harness (Connect, Contextualize, Create, Control) on top of existing legal systems, as Newcode describes its product, rather than a single vendor's fixed AI features.

• Configurable AI harness: Newcode's phrase for a layer that lets a firm choose models, connectors, workflows, and permissions instead of renting a locked feature set.

• Model Context Protocol (MCP): An open standard, introduced by Anthropic on 25 November 2024, for connecting AI applications to external tools and data; Newcode cites 700+ / 750+ MCP connections; Relativity ships a RelativityOne MCP connector.

• Rel Labs: Relativity's innovation and investment program, which joined Newcode's Series A; Relativity says it includes developer support and a strategic investment motion with The LegalTech Fund.

• Nova: Newcode's grounded retrieval and work-product assistant, including DMS search and Wolters Kluwer-powered jurisdictional compare.

• Aurora: Newcode's builder for agentic workflows, with drag-and-drop configuration, guardrails, and deploy-from-one-place.

• System of record: The system that keeps the audit trail — RelativityOne for many e-discovery matters, a practice-management platform or DMS for most small firms — which a harness should read, not replace.

• Agentic retrieval: Newcode's term for a retrieval agent that plans, searches, evaluates, and iterates until context is right, under firm taxonomy and walls.

Frequently asked questions

What is a Newcode AI-native law firm stack?

It is a firm-owned configurable AI harness that connects existing case systems through MCP, retrieves across documents with walls, lets lawyers build agents and work product, and keeps model choice, permissions, and deployment under the firm. It is not a robot lawyer and it is not a Clio replacement.

How much capital did Newcode raise in 2026?

Artificial Lawyer reported a $13.5m Series A on 24 August 2026; both Artificial Lawyer and Legal IT Insider reported about $6.5m seed in March and $20m total 2026 funding. Legal IT Insider noted that Newcode had not itself stated the Series A value in that article.

What does MCP change in a law firm stack?

It replaces one-off connectors with an open protocol so an assistant can talk to mail, DMS, and matter tools through a standard interface. Newcode leads with 700+ to 750+ MCP connections; Relativity uses MCP to orchestrate RelativityOne while keeping that platform as the system of record.

Does Rel Labs backing mean the stack only works on RelativityOne?

No. Rel Labs is Relativity's investment and innovation arm joining a round; Newcode's public connector list is broader (Outlook, iManage, Teams, OneDrive, and 750+ MCPs). Relativity's own MCP page still treats RelativityOne as the e-discovery system of record, which is a parallel product decision, not a lock-in clause in the Newcode announcement.

Should a solo or small firm replace Clio, Smokeball, or MyCase with Newcode?

Not on the evidence in the announcement. Those products are practice-management systems of record with their own AI features; Newcode is pitching the intelligence layer across systems. Pick the matter system first, then decide whether a harness on top is worth the integration work.

What should firms refuse to automate in this stack?

Anything that is a legal conclusion to a client, a filing, or a privilege call without a named reviewer. Relativity's aiR FAQ argues fully autonomous agents without human oversight do not belong in legal tech; the FTC's DoNotPay case shows what happens when "AI lawyer" copy outruns testing.

How do the EU AI Act dates affect a Norway-founded vendor selling into the US?

The Commission dates the Act as in force 1 August 2024 and applicable 2 August 2026, with justice-administration AI on the high-risk list and transparency duties described as taking effect in August 2026. US firms still care because client data, EU offices, and cross-border matters drag those duties into procurement even when the engagement letter is under US law.

Put the harness on work you already run

The Newcode AI-native law firm stack, as of 24 August 2026, is a funded answer to a boring operational question: who owns the intelligence layer on top of the files you already have? Relativity putting Rel Labs into the round, MCP going from an Anthropic spec in November 2024 to 750+ connections on a legal vendor's homepage, and Clio's growing-firm cohort showing revenue moving faster than headcount all point the same way. Own the layer. Keep the record. Put a human on the output.

If you want that orchestration without standing up a legal-only stack, map the same Connect–Contextualize–Create–Control loop on agentic workflows from US Tech Automations. Start at ustechautomations.com. Then put the first connector on the system that already holds the job, the matter, or the chart — not on a chatbot that has never seen it.