Nomic logo
Agents Orchestration Frameworks · Api Tools

Nomic

Nomic is a domain-specific AI agent platform for architecture, engineering and construction firms. It indexes drawings, specs, RFIs and submittals where they already live, then answers, reviews and drafts with citations back to source documents.

Active GDPR compliant Free plan Freemium API available 13+ Verified by Guidaio
Overview

What is Nomic?

Nomic is an AI agent platform built for one industry: architecture, engineering and construction. Its premise is that a firm's expertise sits trapped in drawings, specifications, RFIs, submittals and project systems scattered across SharePoint, Autodesk Construction Cloud, Procore, Egnyte, email and local drives, material that general-purpose AI handles badly. Nomic indexes it where it already sits, through read-only connectors that inherit the permissions of the source system, and runs agents over it that cite their sources.

The product comes in two shapes. Nomic Platform is the turnkey side: a conversational Assistant, Workflows and a workflow builder, plus six documented agents covering code compliance against more than 380 building codes and standards, drawing review, submittal review, firm-wide detail search, project research and BIM/IFC work. The drawing review agent checks every page of a set against the firm's QA/QC standards, produces cited PDF markups, and pushes validated findings back into Autodesk Forma, Fieldwire and Bentley ProjectWise as native issues. The IFC agent reads models without Revit or Navisworks, extracting them into queryable tables for quantity takeoffs, coordination checks and space schedules.

Agent API targets internal AI teams, with four endpoints: Parse turns drawing sets and complex PDFs into structured output with layout, OCR and tables preserved; Extract produces schema-driven structured data; Embed builds multimodal embeddings across text, images, drawings and specs; Datasets governs the collections. Nomic's models are trained on scanned drawings, specs and contracts rather than clean web text, the company's stated reason they cope with 400-sheet sets where foundation models fail.

Eleven connectors have dedicated pages, from SharePoint and Egnyte to Teams, Outlook and Gmail. Delivery is SaaS on AWS, tenant-isolated, with regional hosting for Enterprise customers.

Vendor-claimed results include around 20,000 hours saved per year on information gathering and a 92% cut in repetitive work; engineering group Aurecon is quoted citing a 30% productivity increase. The platform is a recent pivot: Nomic published open-source embedding models until announcing it in November 2025.

What it does

  • Answer project questions across drawings, specs, past RFIs and reports, with citations to the source documents
  • Check drawing sets against 380+ building codes and standards and return specific code references
  • Run a first-pass QA/QC review of a drawing set page by page and deliver cited PDF markups
  • Automate submittal review and generate the project's starting submittal log from the full project manual
  • Search thousands of past drawings to find a detail the firm has already drawn
  • Read IFC and BIM models natively and turn them into queryable tables for takeoffs and coordination checks
  • Draft RFI responses, review comments, proposal sections and technical reports grounded in project documents
Audience

When to use Nomic / When not to

A quick filter to help you decide if Nomic is the right fit.

When to use Nomic

  • Architecture and engineering firms whose project knowledge is scattered across SharePoint, Autodesk Construction Cloud, Procore and Egnyte
  • QA/QC and design review teams checking drawing sets page by page before issue
  • Project managers and engineers who lose hours researching RFIs, submittals and code requirements
  • Internal AI teams that need to parse 400-sheet drawing sets and build their own retrieval on top, through the Agent API
  • Firms of 25 seats or more, and global groups with data residency requirements in the EU, UK, US or Australia

When not to use Nomic

  • Teams outside architecture, engineering and construction: the models, the agents and the 380+ code library are built for drawings and specs
  • Organisations that require on-premise or customer-VPC deployment, which the security page states is not offered
  • Small teams of three to twenty people, since nothing sits between the single-seat Individual plan and a 25-seat Business minimum
  • Developers on the Free or Individual plans, who get no Agent API access
  • Anyone needing a mobile app or a browser extension: Nomic ships as a web platform and an API only
Get started

How to use Nomic

A typical end-to-end flow, from setup to results.

  1. Sign up from the site, where the Free plan needs no credit card, or sign in through your identity provider
  2. Authorise a connector to a file repository: SharePoint, Egnyte, Autodesk Construction Cloud, Bentley ProjectWise, Fieldwire, Google Drive, Box, Dropbox, Teams, Outlook or Gmail
  3. Let Nomic index the files or folders you point it at; existing source-system permissions are inherited, so no access matrix has to be rebuilt
  4. Ask questions in the Assistant and read the answers with their citations back to the source documents
  5. Run an out-of-the-box agent: code compliance, drawing review, submittal review, detail search, project research or BIM/IFC
  6. Review the output yourself, since the terms make human verification the customer's responsibility
  7. Build repeatable Workflows in the workflow builder for the tasks your team runs most often
  8. On Business and Enterprise, configure SAML/OIDC single sign-on through WorkOS with Azure AD or Okta, and assign admin or member roles
  9. Track consumption through usage analytics and the audit log, and top up the AI usage balance when it runs low
  10. For custom tooling, build against the Agent API's Parse, Extract, Embed and Datasets endpoints, documented on docs.nomic.ai
Quick read

Pros & Cons

Pros

  • Genuine domain specialisation: models trained on scanned drawings, specs and contracts rather than clean web text
  • Files stay in place and existing permissions are inherited, which removes the migration and access-rights work most deployments demand
  • Answers carry citations to source documents, the minimum condition in a trade where every claim has to be checkable
  • Unusually precise security page: subprocessors named one by one with their file-access level, AWS regions named, retention spelled out
  • Training on customer data is contractually forbidden, with signed zero-data-retention agreements at Anthropic and Google Cloud Vertex
  • SOC 2 Type II certification and at least annual third-party penetration testing, with reports available under NDA
  • Public pricing with real figures, a free plan without a credit card, and AI usage billed at underlying model rates with no markup

Cons

  • No postal address anywhere on the site, no legal notice and no contact page: only 'headquarters in New York City' on the careers page
  • The Business tier starts at 25 seats, roughly USD 12,000 a year before AI usage, with nothing between it and the single-seat Individual plan
  • Real cost is hard to forecast, since AI usage is billed as a dollar balance on top of seats and the included USD 20 per seat buys only rough volumes
  • The homepage promises deployment 'in your cloud or on your servers' while the security page states that no on-premise or customer-VPC deployment is offered
  • The privacy policy, dated 15 January 2026, still relies on the Privacy Shield frameworks invalidated in 2020, and names neither a DPO nor an Article 27 EU representative
  • EU and UK data residency, like the Agent API itself, sits behind the higher tiers
  • Performance figures are vendor claims with no published methodology, and the ROI table rests on an assumed USD 200 hourly rate
Pricing

Pricing & Plans

Nomic offers a permanent free plan that requires no credit card. The cheapest paid entry point is the Individual plan at USD 20.00 per month. Team pricing begins at USD 40.00 per user per month on the Business plan, subject to a 25-seat minimum, and Enterprise pricing is quoted on request. AI usage is billed separately as a dollar balance charged at the underlying model rates without markup; Business and Enterprise include USD 20 of usage per seat per month, pooled across the organisation.

Free — no charge and no credit card
  • limited agent requests
  • access to the core Platform surfaces
Business — USD 40 per user per month with a 25-seat minimum
  • USD 20 of included AI usage per seat
  • project delivery
  • file integrations (Autodesk
  • Bentley
  • SharePoint and more)
  • centralised billing and administration
  • usage analytics
  • org-wide privacy controls
Enterprise — custom pricing
  • pooled usage commitments
  • invoice or PO billing
  • SCIM seat management
  • audit logs and compliance reporting
  • multi-region data hosting (EU
  • AU
  • US
  • UK)
Prices and plans listed above may evolve. Always check the official pricing page before subscribing.
Trust & Privacy

Data, GDPR & hosting

A consolidated view of how Nomic handles your data.

GDPR overview

Nomic claims broad compliance with major regional data protection regulations and backs it with a signed DPA covering EU GDPR, UK GDPR, CCPA and other US state laws, Brazil's LGPD and the Australian and New Zealand Privacy Acts. Data subject requests are answered within one month, breaches are notified to customers within 24 to 48 hours, and subprocessor changes carry at least 15 days' notice with a documented right to object. Enterprise customers can host data in Frankfurt or London. Two caveats: the DPA published in full is the US State version, the Global DPA being available only on request, and the privacy policy, dated 15 January 2026, still invokes the Privacy Shield and Swiss-U.S. Privacy Shield frameworks invalidated in 2020. No Article 27 EU representative and no DPO is named.

Who owns the data?

Nomic's terms are explicit: every file, content, piece of information and data uploaded by the customer or generated by the service remains the customer's property. Nomic receives only the licence it needs to run the service, for the duration of the order, and customers can export their data through the product while the contract lasts. Administrators can pull user-generated content, usage logs and audit data out as CSV or JSON. Nomic reserves one thing for itself: aggregated, de-identified information that cannot be traced back to a customer, whose intellectual property it keeps. Customers warrant in turn that they hold the rights and consents covering whatever they upload.

Reuse rights

Customers keep the right to use and export their own material without asking Nomic for permission; the licence runs the other way, and is limited to what providing the service requires. Two things are forbidden on the customer side: reverse-engineering the service, and using it to build a competing product or copy a feature for competitive purposes. Feedback and suggestions sent to Nomic, by contrast, may be used freely by Nomic with no obligation in return. On its own side, Nomic states plainly that no customer data is allowed to train any AI model, and that its Anthropic and Google Cloud Vertex inference subprocessors operate under signed zero-data-retention agreements. The data processing addendum adds that Nomic does not sell or share personal information and does not combine it with data received from other sources.

Data retention & training

Retention summary
Nomic separates two retention layers. What it keeps as your processor: user-generated content, meaning conversations, reports, annotations and indexes, for the length of the subscription unless you delete it, plus usage metadata such as session logs, query timestamps and user activity kept for audit and platform operations. What its AI inference subprocessors keep: nothing. Anthropic and Google Cloud Vertex process documents ephemerally under signed zero-data-retention agreements and discard them. On termination you choose between having your data returned in a commercially reasonable format or erased; either way all customer data is deleted within 30 days using secure erasure methods, with written certification available on request. Administrators can export content, usage logs and audit data as CSV or JSON at any point during the contract, and data subject requests are handled within one month.
Trains on customer data
No
Subprocessors disclosed
Yes
DPA available
Yes
GDPR contact

Hosting summary

Nomic runs as a managed platform on AWS, with infrastructure isolated per tenant. Four regions are offered: eu-central-1 in Frankfurt, eu-west-2 in London, us-west-2 in Oregon and ap-southeast-2 in Sydney. The region is chosen at onboarding, and multi-region residency is an Enterprise feature announced in June 2026. Residency covers files stored by Nomic, indexes, cached documents, conversations, usage logs and audit data. The default remains the United States: the published US State DPA states that all primary infrastructure sits on AWS within the United States, and the privacy policy says plainly that Nomic's servers are located in the US. AI inference is the exception, since content leaves Nomic's infrastructure for Anthropic and Google Cloud Vertex over TLS 1.2 or higher, under signed zero-data-retention agreements. Data is encrypted with AES-256 at rest. Nomic states it operates no infrastructure in China and uses no Chinese company as a direct subprocessor.

Hosting countries
🇩🇩 Germany🇦🇺 Australia🇺🇸 United States🇬🇧 United Kingdom
Hosting regions
EUAUUSUK
Availability

Where Nomic works

Country-level availability.

Not available in

Countries under United States embargo: the terms of service forbid access from any US Embargoed country, without naming them
Watch-outs

Things to keep in mind

Risks and trade-offs to weigh before adopting Nomic.

  • Answers look authoritative because they are cited, which makes it easy to stop checking them; the terms of service put human review of every AI output squarely on the customer
  • Delegating drawing review and code checking can erode the very expertise the tool exists to surface: junior staff who never do the manual check never learn to spot what the agent misses
  • Indexing a firm's entire history turns one search box into a single point of exposure; if source-system permissions were already too loose, Nomic will faithfully inherit that looseness and make it findable
  • Customer files are seen by three inference subprocessors, Anthropic, Google Cloud Vertex and Modal Labs, under zero-retention agreements that are contractual promises rather than technical guarantees
  • Cost creeps quietly, because usage is a dollar balance drawn down by chats, indexing and API calls: an enthusiastic team can spend well beyond the seat price without noticing
  • The vendor's own numbers, 20,000 hours a year, 92% less repetitive work, a USD 200 hourly rate in the ROI table, set expectations that no published methodology supports
  • One unattributed testimonial, credited only to a 'Project Manager, Global AEC Firm', sits alongside the sourced Aurecon quotes and deserves more caution than the named ones
Setup

Setup & Integrations

Technical difficulty

Low for the platform. Sign-up is self-service on the free plan, and connecting a repository is an authorisation step rather than a project: no migration, no custom development, and existing source-system permissions are inherited, so no access matrix has to be rebuilt. Nomic advertises setup in minutes. Business and Enterprise add SAML/OIDC single sign-on through WorkOS, plus SCIM on Enterprise, which is routine work for an IT team. The Agent API is another matter and calls for real development against the Parse, Extract, Embed and Datasets endpoints. Nothing is installed locally, and no on-premise deployment is available.

Deployment

Web appAPI

Integrations

SharePoint Egnyte Autodesk Construction Cloud Bentley ProjectWise Fieldwire Procore Autodesk Forma Google Drive Box Dropbox Microsoft Teams Outlook Gmail Okta Azure AD
Company

Behind Nomic

Company name
Nomic, Inc.
Founded
19/12/2021
Country of origin
🇺🇸 United States
UBO
INFORMATION_NOT_FOUND
UBO country
INFORMATION_NOT_FOUND
Domain registrar country
🇺🇸 United States
Support contact

Fundraising

Series A of USD 17 million announced on 13 July 2023, led by Coatue, with Contrary Capital, Betaworks Ventures, SV Angel, Story Ventures and Factorial Capital
No funding round is mentioned anywhere on Nomic's own website; this item comes from press coverage and is flagged for human review

Social

Official links

Resources

All the official URLs gathered for verification and reference.

FAQ

Frequently asked questions

Who is Nomic built for?
Architecture, engineering and construction firms, and the internal AI teams inside them. The platform side serves designers, engineers and project managers; the Agent API serves developers building custom tooling over drawings and specs.
Do my files have to be moved into Nomic?
No. Nomic connects read-only to SharePoint, Egnyte, Autodesk Construction Cloud, Bentley ProjectWise, Fieldwire, Google Drive, Box, Dropbox, Microsoft Teams, Outlook and Gmail, indexes the content in place, and inherits the permissions already configured there. Only directly uploaded files are stored by Nomic itself.
Is my data used to train AI models?
No. The terms of service state that Nomic does not allow any customer data to be used to train any AI model, and the Anthropic and Google Cloud Vertex inference subprocessors work under signed zero-data-retention agreements.
What does Nomic cost?
Free with no credit card; USD 20 per month for Individual; USD 40 per user per month for Business from 25 seats; Enterprise on quotation. AI usage is charged separately at underlying model rates, with USD 20 per seat included on Business and Enterprise.
Where is my data hosted?
On AWS. Enterprise customers pick a region at onboarding: Frankfurt, London, Oregon or Sydney. AI inference is the one point where content leaves Nomic's infrastructure, going to Anthropic and Google Cloud Vertex under zero-retention terms.
Is there an API?
Yes, the Agent API, with Parse, Extract, Embed and Datasets endpoints, documented at docs.nomic.ai. Access is included on the Business and Enterprise plans only.
What security certifications does Nomic hold?
SOC 2 Type II, audited annually, plus at least annual third-party penetration testing. Those reports, along with Anthropic's zero-data-retention certificate, are available under NDA through the trust centre.
How long is my data kept?
User-generated content and usage metadata are retained for the life of the subscription unless you delete them. On termination all customer data is erased within 30 days using secure erasure methods, with written certification of deletion on request.
Can Nomic run on our own servers?
The security page says no: the offering is a Nomic-managed AWS platform, with regional data residency replacing on-premise and customer-VPC deployment. The homepage still advertises deploying 'in your cloud or on your servers', so the point is worth clarifying with sales.
Is there a mobile app?
None was found. Nomic ships as a web platform and an API; no iOS app, Android app or browser extension is listed anywhere on the site.
Conclusion

Should you pick Nomic?

Nomic is one of the more substantiated enterprise AI products a directory is likely to review. The company is identifiable, Nomic, Inc., a Delaware corporation headquartered in New York; three dated legal documents are published in full; the security page names every subprocessor and whether it can see customer files; and the pricing page carries actual numbers rather than a contact form. For an AEC firm the architecture is the argument: files stay in SharePoint, ProjectWise or Autodesk Construction Cloud, permissions are inherited rather than rebuilt, and every answer arrives with a citation, which matters more here than elsewhere because a wrong code reference on a permit set is expensive.

The reservations are specific. No postal address appears anywhere on the site and there is no contact page; the only route in is a demo form. The 25-seat Business minimum leaves mid-sized teams stranded, and AI usage is billed on top of seats as a variable balance. Homepage and security page contradict each other on self-hosting, the latter, more recent and more precise, saying it is not offered. The privacy policy still cites the Privacy Shield frameworks struck down in 2020. Productivity figures throughout the site are vendor claims with no published methodology.

Note too that this is a recent pivot: Nomic published open-source embedding models for years before announcing this AEC platform in November 2025. The engineering pedigree shows in the parsing and retrieval work, but the vertical product itself is young. Firms with the seat count and a real drawing-review or code-compliance burden should evaluate it seriously; smaller practices can start on the free or Individual plan.