Nunc is a managed AI gateway platform launched by Atwood Design, positioned as the "governance layer" between and LLM. It provides core capabilities such as authentication, PII desensitization, policy execution, route distribution, and audit tracking to ensure that regulated organizations meet compliance requirements such as ISO 42001, SOC 2, and GDPR in enterprise-level AI deployment. Nunc is delivered in a managed service model and supports front-end Agent access such as Claude and Claude Code, covering vertical industries such as associations, finance, law, construction/IoT, and HR.

Nunc’s Managed AI Gateway Deep Dive

Core parameters and statistics

Nunc is essentially a "governance agent layer" deployed between the internal boundary of the enterprise and the external LLM. Its core parameters cannot be measured by traditional model parameter quantities or context lengths, but are reflected from the architectural dimension:

Parameter Dimensions	Nunc	Guardrails AI	AWS Bedrock Guardrails	Traditional API Gateway
Core positioning	Operational-level AI governance gateway	Open source AI guardrail framework	Cloud native content security filtering	API routing and current limiting
Connection protocol	MCP, REST, A2A, custom Worker	Only supports LLM call interception	Bedrock native integration	HTTP/REST/gRPC
Compliance Alignment	ISO 42001, SOC 2, NIST AI RMF, EU AI Act	Self-configuration	AWS Compliance Endorsement	No built-in AI compliance
PII processing	Automatic entry desensitization + exit restoration (tokenized two-way mapping)	Rule-based output filtering	Sensitive word/content filtering	Not supported
Policy granularity	Tool-level operation permissions + role-level access control + manual confirmation points	Model output verification rules	Content security policy	IP/URL/current limiting policy
Audit granularity	Full link tracking (request→policy→routing→response→manual)	Model call log	CloudTrail integration	Basic request log
Delivery Model	Hosted Services (Operated by Atwood)	Self-Hosted/Enterprise Hosting	Cloud Service Integration	Self-Deployed/Cloud Managed
Online cycle	~60 days (including management and construction)	Hours to days	Minutes (with AWS)	Days to weeks
Front-end Agent support	Claude, Claude Code, BYO Agent	Universal LLM access	Bedrock model	Not applicable
Applicable organization size	Medium and large regulated enterprises	AI companies with complete technical teams	Enterprises within the AWS ecosystem	All sizes

Interpretation: Nunc is not an independent AI model or SaaS product, but a layer of "infrastructure middleware". It does not directly provide conversational or generative capabilities, but rather enables existing LLMs such as Claude to securely perform tasks on the real systems of regulated enterprises. Compared with traditional API gateways and AI guardrail solutions, Nunc's core differences lie in three dimensions: deep embedded compliance (ISO 42001 controls are directly encoded as platform capabilities), operational-level governance granularity (not only blocking "what is said", but also "what is done"), and fully managed delivery (customers do not need to form an internal AI team). This also explains why Nunc's launch cycle is much longer than that of pure software solutions - what it delivers is not a piece of configuration code, but a set of compliance-verified governance systems.

User and market recognition

Nunc was officially launched by Atwood Design in early 2026 and is currently delivered as a hosted service. The number of unique users or subscriptions has not been publicly disclosed. However, its market position can be assessed from the following dimensions:

• Client background: Atwood’s past service clients include large organizations such as Microsoft, Marriott, Procter & Gamble, AWS, Tradeweb, Exelon, University of Maryland, etc. Its governance framework and design methodology have been verified in enterprise-level projects.
• Industry Coverage: Nunc focuses on covering 6 verticals - Association/Nonprofit, Financial, Legal, Civil/Geospatial, Construction/IoT, HR/Operations - which are characterized by strict regulation, data sensitivity, and lack of in-house AI teams.
• Compliance Endorsement: Nunc is one of the few AI governance platforms that publicly declares to be aligned with ISO 42001 (international standard for AI management systems) and is compatible with SOC 2, NIST AI RMF, EU AI Act and other frameworks, which is a very critical "compliance clearance" certificate in corporate procurement.
• Competitive Positioning: In the enterprise AI governance track, Nunc's direct competitors include Guardrails AI, LangSmith (monitoring layer), and AI Gateways of major cloud vendors (AWS Bedrock Guardrails, Azure AI Content Safety). But the core difference of Nunc is that it is an "operator-grade managed service" rather than a self-service platform - customers do not need to build an in-house AI team to implement it.

Market Recognition Summary: Nunc is currently in the early commercialization stage, and public cases have not yet been disclosed on a large scale. But the need for "regulated enterprise AI governance" it targets is real and urgent - Gartner predicts that by 2027, 60% of enterprise AI deployments will require some form of governance middle layer. Nunc’s first-mover advantage lies in its differentiated positioning of ISO 42001 alignment + design thinking integration, but market share data remains to be verified.

Cost advantage

Nunc uses a fixed pricing + managed service model and does not provide a public price list. The following is an analysis based on its official public information and industry benchmarks:

Dimensions	Nunc	Guardrails AI	AWS Bedrock Guardrails	Build your own solution
Pricing model	Fixed quote (per Pilot/project)	Open source free + Enterprise subscription	Billed by Guardrail calls	Internal team hours + infrastructure
C-side pricing	Not applicable	Not applicable	Not applicable	Not applicable
Developer/API	No separate API subscription available	Free for open source version; $2,500+/month for enterprise version	Billed per Guardrail processing unit	Internal development costs
Enterprise/Private	Fixed Pilot quotation (about 60 days) → Monthly access expansion	Enterprise plan requires business communication	Pay-as-you-go billing within the AWS ecosystem	Full-time team + cloud infrastructure
Hidden costs	Requires continuous managed operations (Atwood management)	Labor costs of self-operation and maintenance/self-integration	AWS binding and Vendor Lock-in	Recruitment, training, staff turnover risks
Compliance certification cost	Built-in ISO 42001/SOC 2	Self-certification required	AWS compliance endorsement	Build a compliance system from scratch

FREE FACT: Nunc does not have a free tier or free trial. Its delivery model requires customers to first conduct a pilot with a fixed quotation (about 60 days), and then expand the access control on a monthly basis after passing the verification. This means a budget commitment is required before value can be generated, making it suitable for regulated organizations with existing AI budgets, but not suitable for individual developers or small and micro businesses to test the waters.

C-side cost: Nunc is not for C-side users, and individuals cannot subscribe directly.

API/Developer Cost: Nunc is not sold separately as an API and does not provide options for billing by token or by call volume. Developers cannot call Nunc directly like they can with the OpenAI API.

Enterprise Cost Structure:

• Pilot Phase: fixed price, fixed scope, approximately 60 days delivery, covering governance access to a high-value workflow.
• Expansion Phase: Access control is expanded on a monthly basis according to the "Department System" model. Each access control has a fixed scope and quotation.
• Operation Phase: Continuous hosting operations adopt the retainer model and are priced based on value to customers.

Comparative Advantages: Compared with self-built AI governance solutions, Nunc can compress the time from "zero to compliant online" from 6-12 months to about 60 days, and avoids hidden sunk costs such as recruiting an AI team, purchasing infrastructure, and building a compliance system. For mid-to-large organizations with heavy annual compliance audit pressures but missing AI teams, this time difference in itself is a significant ROI.

Main functions

Nunc's core capabilities revolve around the role of "AI Governance Gateway". The following are the key functions arranged in workflow order:

• Authentication & Authorization: All requests entering Nunc must pass identity authentication, and the system determines which tools and data sources the identity can access based on policies. Supports integration with existing IdPs (Identity Providers) to avoid spreading the management of credentials across multiple Agents.

• PII Stripping: Automatically identify and strip personally identifiable information (name, address, phone, email, etc.) before the request reaches LLM, and replace it with an irreversible token placeholder (such as [DONOR_07]). LLM performs inference on the desensitized data, returns the results, and then restores the token to real data locally. This way the original sensitive data never leaves corporate boundaries.

• Policy Engine: A configurable set of governance rules that defines "who can perform what actions on what tools." The policy covers data access scope, operation type (read-only/write/delete), time window, approval level and other dimensions. High-risk operations (such as payment, deletion, and publishing content) are automatically suspended and await manual confirmation.

• Routing: After unifying the portal, Nunc will route the request to the correct back-end system - CRM, financial system, document library, knowledge base, etc. Supports multiple protocols such as MCP, REST, and A2A, allowing front-end agents such as Claude to access multiple backends through a single gateway.

• Full Audit Trail: The complete path of each request from entering the gateway to returning the result is recorded, including: who initiated the request, which policy checks were passed, which back-end tools were called, what results were returned, and whether there was any manual confirmation. Audit logs can be used for compliance review and follow-up.

• Prompt Injection Defense: Agents may be hijacked by implicit instructions when processing untrusted content (email PDFs, web pages, calendar invitations). Nunc screens and sterilizes input at the gateway layer, and at the same time limits the scope of the tool so that unauthorized operations such as "export all contacts" cannot be performed.

• Observability & Monitoring: The production environment continuously monitors the quality, delay, and abnormal patterns of LLM calls, and provides dashboards and alarm mechanisms. Monitoring and improvements required by ISO 42001 Clause 9 are implemented here.

• Memory & Context Management (Memory & Context): Supports cross-session persistence and partitioned memory, allowing the Agent to maintain context continuity in long-term tasks. Memories are isolated by organizational boundaries and are not leaked across customers.

[Expert View]: Nunc’s “hidden linkage” lies in its ability to integrate multiple pain points that were originally scattered in enterprise AI deployment—identity management, data desensitization, policy control, compliance auditing, and security protection—into a unified gateway layer. Usually enterprises need to purchase 3-5 independent products (SSO solution + data desensitization tool + API gateway + log audit system + LLM security scanner) to piece together similar capabilities, but Nunc natively couples them with AI workflows, eliminating the friction of cross-system switching. More importantly, it changes "compliance" from post-document review to pre-architecture built-in - every call has its own compliance attributes.

Model and version evolution

As an enterprise-level governance platform, Nunc's version evolution rhythm is different from that of SaaS/API products, and it is more focused on the iteration of service capabilities and compliance coverage. The following is the version history based on public information:

Version	Time	Key changes	Delivery form
Nunc Beta (v0.9)	~2025-Q4	Internal beta phase, for selected partners. Preliminary verification of core governance gateway functions, supporting MCP connection and basic PII desensitization	Limited internal testing
Nunc 1.0	~2026-Q1	Officially released. Added policy engine, full-link audit, multiple protocol support (REST/A2A), ISO 42001 aligned with 6 major vertical industry solutions	Hosting services
Nunc 1.x (speculated)	~2026-Q2/Q3	Expected: Extended Agent compatibility (GPT/Gemini), added industry compliance package, launched management console Alpha	Managed Services

Version Note: Atwood officially does not release version numbers in the traditional SaaS way, and the above version divisions are based on inferences from publicly available information. The actual evolution of Nunc is more likely to be measured by "service capability level" rather than "software version number" - such as adding a compliance template for a certain vertical industry, expanding a certain MCP connector, upgrading the audit report format, etc. This is consistent with enterprise-level service delivery habits: customers purchase continuously improving governance capabilities rather than regularly updated software licenses.

Comparison of version strategies with competing products:

Dimensions	Nunc	Guardrails AI	AWS Bedrock Guardrails
Version visibility	Undisclosed version number, iterate at service capability level	Open source version number + enterprise version	AWS service version (internal management)
Update frequency	Continuous delivery (managed service)	Open source community quarterly releases	AWS regional grayscale releases
User control	None (Managed by Atwood)	Pinned version	Lockable API version
Backwards Compatible	Guaranteed by Atwood	Maintained by the open source community	AWS Compatibility Commitment

Iteration direction prediction:

• More front-end Agent support (not only Claude, but may be extended to ChatGPT, Gemini, etc.) - This will directly affect the expansion space and bargaining power of existing customers.
• Industry-specific compliance packages (e.g., pre-built policy templates for GDPR, HIPAA, PCI-DSS) – This is a key lever to increase penetration in vertical industries.
• Self-service management console (currently operated by Atwood) - If launched, it will significantly lower the adoption barrier for enterprises with existing AI teams.
• A more fine-grained RBAC/ABAC permission model - to meet the differentiated management needs of multiple departments and roles within large organizations.

Technical advantages

Nunc's technical architecture revolves around the core abstraction of "governance gateway". Its advantages do not come from model capabilities, but from architectural design:

Tool type classification determination

The core delivery form of Nunc is [Agent/MCP/Automation Tool], because it is essentially a governance gateway layer that carries the interaction between AI Agent and enterprise systems. But it also has the attributes of [Productivity/Business-side Application] - it is delivered as a managed service (Managed Service) rather than SDK/API, and what customers purchase is a "governed AI workflow" rather than a simple software component. The following in-depth analysis focuses on its technical mechanism and engineering implementation details as Agent governance infrastructure.

Architecture link: LLM → Nunc Gateway → Enterprise System

The true position of Nunc in the workflow can be represented by the following link:

[Claude / Claude Code / BYO Agent] ←→ [Nunc Governance Gateway]
                                              │
                    ┌──────────────────────────┼──────────────────────────┐
                    │ │ │
               [Authentication layer] [Policy engine] [Audit layer]
          (IdP/SSO integration) (RBAC/ABAC/manual confirmation points) (Full link logs)
                    │ │ │
                    └────────────────────────────────────────────────────────────────┘
                                              │
                    ┌──────────────────────────┼──────────────────────────┐
                    │ │ │
               [CRM/ERP] [Documentation/Knowledge Base] [Other Systems]
           (Salesforce/SAP) (SharePoint/Confluence) (REST/MCP/A2A)

Control flow: The user issues a natural language request through Claude → Nunc gateway interception → Identity authentication → Desensitization of sensitive data → Policy engine determines permissions → Route to the target system → Return results → PII restoration → Audit writing → Present to the user.

Data backflow: The structured data of the enterprise system enters the LLM context after being desensitized, and the inference results are returned to the user after being restored by PII. The original data does not leave the enterprise boundary during the entire process.

Detailed explanation of core technology advantages

1. Gateway architecture vs. point-to-point integration

A common pattern for traditional enterprise AI deployments is to establish separate connections for each LLM to each backend system (Claude→Salesforce, Claude→SAP, Claude→SharePoint…), with each connection requiring separate management of credentials, policies, and audits. When the number of systems in this point-to-point model increases, the security surface and operation and maintenance complexity explode in O(n²).

Nunc adopts the hub-and-spoke gateway architecture: all Agent → system communications are forced to pass through the unified gateway, and authentication, desensitization, policy, and auditing are completed here at once. The complexity is reduced from O(n²) to O(n). To add a new back-end system, you only need to register a route at the gateway without modifying each Agent.

2. Tokenization design for PII desensitization

Nunc's PII processing is not a simple "detect → replace → discard", but uses a tokenization + two-way mapping mechanism. Sensitive data is identified and replaced with meaningless placeholders (such as [DONOR_07]) at the entry, LLM operates on the desensitized data, and after returning the results, it is restored locally through the security mapping table. This is more secure than calling the LLM provider's sensitive data filtering API directly because the raw data never leaves the enterprise-controlled network perimeter. The mapping table is stored encrypted in memory and is automatically destroyed after the session ends.

3. The "confirmation point" mechanism of the policy engine

For irreversible operations (payment, deletion, batch modification, external release), Nunc's policy engine supports setting Human-in-the-loop confirmation point. Models can plan and propose actions, but actual execution requires human approval from designated roles. This is in line with the "Four-Eyes Principle" in the financial field and is a key safety valve for the implementation of enterprise-level AI. The confirmation point can be configured as: approval is always required, approval is required above a threshold (such as amount > $10,000), approval is required during a specific period, etc.

4. ISO 42001 deeply embedded

Nunc is not just "compatible" with the ISO 42001 standard, but encodes its controls (Clause 6-10, Annex A) directly into the platform's capability units. For example:

• Clause 6 (Planning) → Rule definition interface of the policy engine
• Clause 8 (Operation) → Request processing pipeline
• Clause 9 (Performance Evaluation) → Audit log and monitoring dashboard
• Clause 10 (Improvement) → Feedback loop and model behavior tuning

This "compliance-as-code" design reduces clients' audit preparation time from months to days.

5. Structured defense against prompt injection

Nunc does not rely on a simple keyword blacklist to defend against prompt injection, but uses tool scope constraints: the operation parameters of each tool are strictly limited (for example, the "Export Invoice" tool can only output aggregate data of a specific date range, but cannot output all records). Even if the injection instructions try to bypass, the available operation space of the tool is exceeded. This is more robust at the architectural level.

Engineering Pitfall Guide

Based on the architectural characteristics of Nunc, the following engineering issues need to be paid attention to in actual deployment and operation:

Pitfall 1: Deadly loop and Token inflation control Agents may fall into an endless loop of "planning→execution→failure→replanning" when executing multi-step tasks, leading to an explosion in token consumption and out-of-control API fees.

• Solution: Set max_steps (maximum number of steps) and timeout threshold for each Agent task in Nunc's policy engine. It is recommended that the Pilot phase start with an upper limit of 10 steps and gradually adjust after observing the task completion rate. Combined with the "repeated action detection" in the audit log, if the same tool is called more than three times in a row and the result remains unchanged, it will automatically interrupt and notify the administrator.

Pitfall 2: Enterprise system context overload When Agent processes complex records in large ERP/CRM, it is possible to stuff entire customer files or complete financial statements into the LLM context, causing the context window to overflow or the quality of inference to degrade.

• Solution: Implement a "context clipping" strategy at Nunc's gateway layer - perform paginated summaries of system responses that exceed a threshold, and only pass structured summaries to LLM instead of the complete raw data. The max_context_size parameter (unit: Token) can be configured through Nunc's MCP connector, and the excess will automatically trigger the summary pipeline.

Pitfall 3: Security management of irreversible operations AI Agent misoperations (such as bulk deletion of customer records, publishing unaudited financial reports) are unacceptable in regulated organizations.

• Solution: Nunc's policy engine supports "dry-run mode" - all write operations are executed in the Dry-run environment by default and a difference report is generated, and then executed in production after manual confirmation. It is recommended to turn on the confirm flag for all tools involving deletion, modification, and payment during the pilot phase, and then gradually relax it after the Agent's behavior pattern stabilizes. At the same time, set up an "operation rollback plan" to ensure that a snapshot is automatically created before each modification.

How to use

Nunc is not a self-service product, and its usage process follows the service delivery path of "Consultation → Pilot → Expansion → Operation". The following are the access instructions for each entrance:

Entrance	Applicable roles	Access method	Prerequisites
Governed AI Pilot	First-time enterprise customers	Submit requirements → Atwood assessment → Fixed quote → 60-day delivery	Clear high-value workflow candidates
Department System	Expansion stage after Pilot verification	Monthly access control, enter the next stage after customer authorization	Pilot has been accepted and passed
Operate & Expand	Continuous operation of customers	Retainer mode, Atwood continuous operation and optimization	The system is online and running
BYO Agent (bring your own Agent)	Enterprises that already have AI tools	Point Claude/Claude Code, etc. to Nunc gateway	Enterprise-level Claude subscription

Typical access steps (Pilot mode):

1. Discovery: Conduct a workshop with the Atwood team to identify the AI workflow with the highest pain points in the enterprise and sort out the systems and data sources involved. This is not a sales pitch, but a joint demand review.

2. Scope & Price: Solidify Pilot into a fixed scope and fixed price, and launch it only after customer approval. No hidden fees.

3. Build: The Atwood team uses AI to enhance its own development process and quickly build production-level governance connections. Involving: gateway deployment, system docking, policy configuration, PII desensitization rule setting, and audit pipeline construction.

4. Gate & Ship: Monthly milestone gate control. After the customer reviews the deliverables and signs for approval, the system is officially launched. Atwood continues to operate.

Integration example with Claude Code:

# Call Claude Code through Nunc gateway
# Contextual variable points to Nunc endpoint instead of calling Anthropic API directly
export CLAUDE_API_ENDPOINT=https://nunc-gateway.atwood.design/v1
export CLAUDE_API_KEY=<nunc-issued-key>

# Each call to Claude Code automatically passes through Nunc:
# 1. Identity authentication → 2. PII desensitization → 3. Policy check → 4. Routing to Anthropic
# → 5. Return results → 6. PII restoration → 7. Audit write

Note: The above code is a schematic configuration example. The actual access parameters are subject to the Pilot access document provided by Atwood. Nunc does not provide a public API Key self-service application portal.

Product Pricing

Nunc’s pricing model is fundamentally different from API metering or SaaS subscriptions for traditional AI tools:

• Pricing philosophy: fixed price, fixed scope, results-oriented. Not billed based on Token consumption or API calls.
• Pilot Phase: One-time fixed quote covering approximately 60 days of governance gateway construction and first workflow go live. The price depends on factors such as workflow complexity, number of connected systems, compliance requirements, etc., and needs to be determined through business communication.
• Expansion Phase: Promoted by Monthly Milestone Gates, each gate has a fixed scope and fixed additional quotation. Clients sign off and authorize the next phase after each access control, without having to commit the entire budget upfront.
• Operation Phase: Continuous hosting adopts the Retainer (scheduled) method, and the price fluctuates depending on the coverage and service depth.

Pricing Transparency: Nunc’s official website does not disclose prices. This is consistent with the purchasing habits of the target customer group (medium-to-large regulated enterprises) - these customers usually require customized quotes rather than self-service subscriptions. Prospective agencies are advised to contact Atwood directly for a Pilot scope assessment.

Basis for judging cost-effectiveness: For regulated companies with annual revenue of more than US$100 million, Nunc's Pilot quotation will most likely fall within the reasonable range of "enterprise software projects". A reference comparison is: self-building an AI governance layer with the same capabilities requires 3-5 full-time engineers + security compliance experts to work for 6-12 months, plus cloud infrastructure and certification audit fees, the total investment is usually US$500,000-$2 million. Nunc's Pilot offer offers significant value for money if it's below the lower end of that range.

Application scenarios

Nunc's application scenarios focus on "high-value AI workflows in regulated enterprises". The following are three typical scenarios:

Scenario One: Automation of Board Materials for Associations/Non-Profit Organizations

Associations usually use multiple systems such as Salesforce NPSP, Sage Intacct, Cvent, and Bill.com. The preparation of board materials requires the extraction of donation data, financial reports, activity statistics and other information across systems, which takes 2-3 weeks manually and is error-prone. Nunc connects Claude to these systems via a unified gateway:

• Automatically extract donation records and member data from Salesforce (PII is desensitized and passed to LLM)
• Get financial summaries from Sage Intacct and event engagement data from Cvent
• Claude generates draft board materials with traceable citation sources
• Sensitive data (large-amount donor information) remains tokenized in the results and is only visible to authorized personnel

Actual benefits: The material preparation cycle is shortened from 2-3 weeks to 2-3 days, and each generated reference can be traced back to the source system, avoiding the risk of "AI fabricated data".

Scenario 2: Automatic writing of compliance reports for financial institutions

Banks and insurance companies are required to submit monthly/quarterly compliance reports to regulators, which involve obtaining and aggregating data from core banking systems, transaction monitoring systems, and risk management systems. In the traditional way, the compliance team needs to manually capture data, cross-verify, and write report instructions one by one.

Nunc’s governance gateway ensures:

• All data requests go through the policy engine to ensure that analysts can only access data within their scope of authority
• PII is stripped before being transferred to LLM and customer information is not leaked
• The audit log completely records each data extraction path, which can prove the data source to regulatory agencies
• High-risk operations (such as modifying financial numbers in reports) require manual confirmation by the compliance officer

Actual Benefits: Compliance report preparation time reduced from 5-7 days to less than 1 day, audit preparation time close to zero (because all actions are logged).

Scenario 3: Knowledge retrieval and contract drafting in the legal industry

Law firms deal with a lot of precedent searching, contract review, and document drafting. The problem of "illusion" with traditional AI tools is unacceptable in legal scenarios - a fictitious case citation could lead to professional liability.

Nunc’s governance layer provides:

• Restrict LLM to retrieve information only from moderated legal knowledge bases (not the public web)
• Each legal citation comes with a traceable source document ID
• Contract modification suggestions are subject to strategic review to avoid overstepping the authority to modify key clauses
• Confidential client information is desensitized before leaving the firm network

Actual Benefits: Increase search and drafting efficiency by 3-5 times for junior lawyers and legal assistants, while controlling the legal risks of AI illusions at the architectural level.

Applicable people

Nunc's target user group is very focused and not suitable for the mass market:

• Executives and IT decision-makers in regulated enterprises: CISOs, CIOs, CAIOs (Chief AI Officers) – They need to safely introduce AI capabilities without building an in-house AI team. Nunc's "managed service" model exactly solves the pain point of "wanting to use AI but no one to operate and maintain it".
• Compliance & Risk Management Team: Responsible for ensuring that enterprise AI usage meets industry regulatory requirements (SOX, HIPAA, GDPR, PCI-DSS, etc.). Nunc’s built-in audit logs and ISO 42001 alignment significantly reduce compliance review burdens.
• Business-critical users in vertical industries: Directors of Association Operations, Financial Compliance Analysts, Law Firm Managing Partners – who don’t care about technical details but need AI to do specific jobs (board reporting, compliance reporting, contract reviews) within strict security boundaries.

Not suitable for the crowd:

• Individual Developers/Independent Entrepreneurs: Nunc does not have a free tier API subscription or personal version. The minimum delivery unit is enterprise Pilot, which is not suitable for individuals to test the water.
• Large technology companies with existing mature AI governance teams: If there is already an internal AI governance/security team of more than 5 people, building a self-built gateway may be more cost-effective than purchasing Nunc.
• Scenarios that require highly customized model fine-tuning: Nunc provides a governance layer rather than a model layer, and does not involve model fine-tuning or training. If the core requirement is a custom model rather than secure access, look for a MaaS platform.
• Budget-sensitive small and medium-sized enterprises: Nunc’s Pilot model requires upfront budget investment and does not provide standardized monthly subscription options. The threshold is higher for small businesses with tight cash flow.

Summary and Outlook

Core Competencies: Nunc's core value proposition is not "better models" but "making existing models run safely in regulated enterprises". It encodes the ISO 42001 compliance framework directly into platform capabilities, providing a "zero to compliant go-live" managed path for organizations lacking an AI team. In the AI ​​governance track, there is currently no direct competitor for this design and governance integration (Design × Compliance × Operation) positioning.

Current Limitations:

• Supports only managed delivery, does not offer self-hosted or hybrid deployment options, and may not be flexible enough for organizations with extremely strict data sovereignty requirements.
• Currently, the only front-end agents explicitly supported are Claude/Claude Code, and the access to other models (GPT, Gemini, etc.) has not been publicly stated.
• The pricing is opaque and the procurement process is heavy (at least business communication → Pilot evaluation → 60-day delivery required), which is not suitable for quick verification or short-term projects.
• There are limited public customer cases, and the depth and breadth of industry coverage still needs more verification.

Procurement/Adoption Risk Assessment:

• Pilot first: It is recommended to launch Pilot from a single high-value workflow to verify the actual effect of Nunc in your own IT environment and avoid expanding the commitment before the first results are available.
• Focus on exit costs: Nunc uses open protocols such as MCP and REST. In theory, there is no strong Vendor Lock-in, but the actual migration relies on the collaboration of the Atwood team. It is recommended to clarify the agreement standards and data export terms in the contract.
• Compliance Localization: ISO 42001 is an international standard, but if an enterprise is mainly subject to specific regulations such as GDPR, HIPAA, and Class A Protection 2.0, it needs to be verified during the Pilot stage whether Nunc's policy engine can cover the specific control items of these regulations.
• Long-term observation: Will Atwood open up a certain degree of self-service capabilities (management console, self-service policy configuration)? Will Nunc support more LLM backends (e.g. OpenAI, Gemini)? These will directly affect its scalability and bargaining power.

Related tools: CrewAI, LangChain

Version Info

• Nunc 1.0 ：Nunc is officially released as Atwood's core governance platform, supporting MCP, REST, and A2A protocol connections, integrating PII desensitization, policy engine, audit logs, and ISO 42001 alignment. Delivered in a managed service model, covering 6 vertical industries. （~2026-Q1）
• Nunc Beta ：Nunc is an internal beta version, open to selected partners, and is in the core governance gateway functional verification stage. （~2025-Q4）