Fermi Dev
Fermi Dev is an AI Operational Brain for regulated, high-stakes enterprises: it connects CRM, ERP, databases and internal APIs into one live operating model, then lets AI agents execute work under simulation, human approval and audit.
What is Fermi Dev?
Fermi Dev, published by Fermi Development, Inc., presents itself as the AI Operational Brain for modern enterprises. Its premise is that companies fail not for lack of tools but from fragmentation: work is scattered across 20+ systems, tribal knowledge sits with a handful of people, documentation lags behind reality, and AI projects stall because nothing holds the live operational context they would need. The product rests on three pieces. The Brain is a real-time digital twin of the company, capturing entities (customers, orders, products, inventory, teams, resources), relationships (system dependencies, approval chains, hierarchies), policies (business rules, compliance, governance) and KPIs. It is assembled through AI onboarding - ingesting documentation, connecting to live systems, then learning continuously from team feedback so it does not go stale. It answers questions in natural language, runs automations and recommends actions. The Conductor governs change through a Simulate, Approve, Apply loop: it models the impact of a proposed change on workflows, teams, systems and KPIs, puts it in front of a human with that context, then applies it with rollback, audit and supervision. AI Agents carry out execution, drawn from a library of pre-built agents or built with a custom agent builder, on three stated principles - human-in-the-loop, a transparent chain of reasoning, and a complete audit log - sharing context through the Brain. Fermi positions the product against RPA, which scripts the surface and breaks when a process changes; against point-solution AI, one tool per question; and against general-purpose LLMs, which hallucinate without live operational context. The industries it targets are pharmaceutical manufacturing - the public demo organization, Fermi Pharmaceuticals, covers SOPs, batch records and quality workflows - plus financial services and logistics, whose demo organizations are still in progress, with healthcare and energy also cited. Security claims are specific: SOC 2 Type II, AES-256 encryption in transit and at rest, encrypted credential storage, SSO/SAML, role-based access control, exhaustive audit logs that cover agent actions, and deletion within 30 days of disconnection. Fermi says most customers are operational within days rather than weeks, and serves organizations from 50-person startups to the Fortune 500. The vendor also advertises 60% lower operating costs and 35% higher throughput on its homepage, figures published with no methodology and no source.
What it does
- Connect CRM, ERP, databases and internal APIs into a single operating model
- Build a living map of the business automatically: entities, relationships, policies and KPIs
- Query operations in natural language and get answers grounded in live systems
- Simulate the impact of a workflow change before applying it
- Route the change through human approval, then apply it with rollback and an audit log
- Deploy AI agents that execute work inside existing systems
- Generate operational reports and dashboards
When to use Fermi Dev / When not to
A quick filter to help you decide if Fermi Dev is the right fit.
When to use Fermi Dev
- Operations leaders in pharmaceutical manufacturing, who get a pre-configured demo organization built around SOPs, batch records and quality workflows.
- IT and engineering executives in mid-market and enterprise companies whose teams are, in Fermi's words, drowning in coordination across 20+ disconnected systems.
- Financial services, logistics and supply chain teams, the two industries whose demo organizations Fermi lists as in progress alongside pharma.
- Regulated, high-stakes organizations - healthcare and energy included - that need human approval, rollback and audit trails on every automated action.
- Companies from 50-person startups to the Fortune 500 that want AI agents which execute work inside existing systems, not agents that only answer questions.
When not to use Fermi Dev
- Teams that need a published price or a self-service purchase: there is no pricing page and every route runs through a sales demo.
- Anyone expecting a mobile app or a browser extension - Fermi ships as a web app and an API only.
- Non-English-speaking teams: the interface and the product content are English only.
- Individuals and personal users: the Terms restrict the service to internal business purposes and to people aged 18 or over.
- Organizations with no real systems to connect, since the value comes from plugging into live CRM, ERP, databases and internal APIs.
How to use Fermi Dev
A typical end-to-end flow, from setup to results.
- Look before you talk to anyone: the "Explore Product" button opens the Fermi Pharmaceuticals demo organization at demo.fermi.dev, advertised as free to explore with no sign-up needed.
- Book a demo from the contact page, which embeds a 30-minute Cal.com slot; Fermi says it typically responds within one hour.
- Go through the three announced steps: a discovery call on your workflows with no pitch, a demo built around your use cases, then an implementation plan.
- Start the Brain's onboarding, stage one - Deep-Dive Discovery: conversations with your teams and ingestion of your documentation.
- Stage two - Live System Mapping: connect CRM, ERP, databases and tools so Fermi can map flows and dependencies.
- Connect the sources: Google (Drive, Sheets, Sign-In), Microsoft 365 read-only through the Graph API (Outlook, SharePoint, Teams, Dynamics 365), SQL databases, MongoDB and API endpoints - credentials are stored encrypted.
- Stage three - Always-On Adaptation: the Brain keeps learning and workflows are updated as the business moves.
- Use it day to day: ask questions in natural language, simulate a change, have it approved, then apply it.
- Deploy agents from the pre-built library or build your own, and follow what they do in the audit log.
- Developer route: create an API key, create a service actor, exchange it for a token, then call the Analytics, Brain, Chat & Sessions, Custom Ontologies and Data Connectors endpoints, with Python and TypeScript SDKs.
Pros & Cons
Pros
- The Simulate, Approve, Apply loop puts human validation and rollback in writing before a change reaches production - an uncommon commitment in this segment.
- The security posture is named rather than hinted at: SOC 2 Type II, AES-256, SSO/SAML, role-based access control and audit logs that also cover the actions of agents.
- Public and complete API documentation at docs.fermi.dev, with Python and TypeScript SDKs.
- A pharmaceutical demo organization can be explored at demo.fermi.dev with no sign-up, so the product can be judged before any sales conversation.
- Subprocessors are named in the privacy policy - Google Cloud, AWS, Microsoft, Stripe and the AI providers - instead of being left generic.
- Explicit commitments: no sale of personal or organizational data, and third-party AI providers are barred from training their models on customer data.
- Data handling details that are rare at this size: Microsoft 365 connectors strictly read-only and revocable, deletion within 30 days of disconnection, and a designated DPO at dpo@fermi.dev with a 30-day response commitment.
Cons
- No public pricing whatsoever: no pricing page, no amount, no tier - the JSON-LD markup says only "Contact for enterprise pricing".
- No documented permanent free plan, and free trials are only a possibility: the Terms (4.3) say Fermi "may offer free trial periods".
- No postal address is published on any page - there is no legal notice, and the Terms and privacy policy give email addresses only.
- No customer DPA is published or offered, no opt-out from model training is documented, GDPR compliance is never claimed explicitly and no Article 27 representative is designated.
- No hosting country or region is specified anywhere; only Google Cloud Platform and Amazon Web Services are named.
- The site is rendered entirely in JavaScript with no static content, so nothing is readable without JS execution; the product is English only, with no mobile app.
- A young company - domain registered on 31/12/2024 - whose financial services and logistics demo organizations are still in progress, whose homepage statistics (60% cost reduction, 35% throughput) come with no methodology or source, and whose footer social links point to the Twitter, LinkedIn and GitHub home pages rather than to brand accounts.
Pricing & Plans
No price is published. Fermi Dev has no pricing page, and no amount appears anywhere on the site or in its llms.txt and llms-full.txt files; the JSON-LD markup carries a placeholder price of 0 USD with the note "Contact for enterprise pricing", which is not a real rate. There is no documented permanent free plan - the only free access is the pharmaceutical demo organization, advertised as free to explore with no sign-up needed. The Terms (4.1) refer to several subscription plans whose features, limitations and pricing are said to be described on the website, although no such page exists. Where a subscription applies, it is billed in advance on a recurring monthly or annual basis and fees are non-refundable except where the law requires otherwise (4.2); free trial periods may be offered and convert automatically into a paid subscription unless cancelled (4.3), and cancellation takes effect at the end of the current billing period (13.2). Payments are processed by Stripe. A figure can therefore only be obtained by contacting the vendor's sales team.
- no plan name
- no tier and no amount appears on any page of the site.
- The Terms refer to various subscription plans billed monthly or annually
- without naming or listing any of them.
- The only way to obtain a rate card is the sales appointment booked from the contact page.
Data, GDPR & hosting
A consolidated view of how Fermi Dev handles your data.
GDPR overview
Fermi Dev never claims GDPR compliance in so many words: "GDPR" appears only as a meta keyword on the privacy page, and no EU representative under Article 27 is designated. What does exist is concrete. The privacy policy and the Terms both took effect on June 11, 2026. Section 8 grants access, rectification, erasure, objection and restriction, portability and withdrawal of consent - qualified by "depending on your location" rather than granted universally. A data protection officer is designated and reachable at dpo@fermi.dev, with a stated response time of 30 days (8.3). Section 10 acknowledges international transfers under "appropriate safeguards" that are never named - no standard contractual clauses are cited. The service is closed to anyone under 18 (11). Governing law is Delaware, with AAA arbitration (Terms, 15). No customer DPA is published or offered.
Who owns the data?
Section 6.1 of the Terms is explicit: you retain all ownership rights in the data you upload, input or submit, defined as "Your Data". Fermi Development, Inc. receives only a limited license to process, store and analyze it for the purpose of operating the service, and the privacy policy (4.3) confirms that Fermi acts as a data processor on your behalf. Personal and organizational data is not sold (5). Microsoft 365 connectors are read-only - Fermi states it does not write to, modify or delete any data in the connected source. Conversely, the service itself, its software, algorithms and designs remain Fermi's property under section 7 of the Terms.
Reuse rights
Because Your Data remains yours, you can export and reuse it without asking Fermi's permission; Fermi's own rights go no further than running the service. Section 3 of the privacy policy lists the purposes: authentication, organization onboarding, building the intelligence layer, analytics, agent deployment, reports and dashboards, billing, support, security and legal obligations. Data pulled from connected sources is used for no purpose other than providing the service (4.3). Named third parties touch it: Google Cloud Platform, Amazon Web Services, Microsoft through the Graph API, Stripe for payments, analytics providers, and the AI providers OpenAI, Anthropic and Google AI - which, section 4.4 states, are not allowed to use your data to train their models. Data from Google APIs is additionally governed by the Limited Use requirements of the Google API Services User Data Policy. One clause deserves attention: section 4.2 says the system learns from patterns in your organization's data to improve performance - the only mention of learning on customer data, and worth pinning down contractually. Browser data, IP address, time zone and cookies are collected automatically (2.9).
Data retention & training
Hosting summary
Fermi names its hosting providers but not their locations. The privacy policy (5) identifies Google Cloud Platform and Amazon Web Services as infrastructure providers; no hosting country and no region is stated anywhere on the site, which is why both are left empty in this record. Section 10 accepts that data may be transferred outside your own jurisdiction, under "appropriate safeguards" that are never named - no standard contractual clauses are cited. The measures described in section 6 are more specific: end-to-end encryption in transit and at rest, AES-256, SSL/TLS, firewalls and intrusion detection, secure facilities with restricted access, regular security audits and penetration testing, role-based access control, secure credential management, and infrastructure certified SOC 2 Type 2. The policy also carries the usual caveat that no method of transmission over the internet is 100% secure. For any organization with data residency requirements, the hosting jurisdiction is the missing piece and has to be established directly with the vendor.
Things to keep in mind
Risks and trade-offs to weigh before adopting Fermi Dev.
- No public pricing: budgeting is impossible without going through the sales team, and the Terms (4.1) refer to a pricing page that does not exist on the site.
- Section 4.2 of the privacy policy says the system learns from patterns in your organization's data to improve performance, which sits alongside the promise that third-party AI providers cannot train on that data - reconcile the two in the contract before signing.
- No customer DPA is published or offered, no Article 27 EU representative is designated and GDPR compliance is never claimed explicitly - a gap to close before any European deployment.
- No hosting country or region is specified, and international transfers rely on "appropriate safeguards" that are never named.
- Setting the tool up means handing over database credentials and Microsoft and Google tokens: a risk surface to arbitrate with your security team, even though the Microsoft 365 connectors are read-only.
- The Terms allow Fermi to modify, suspend or discontinue the service at any time without notice (14), and cap liability at the amounts paid over the preceding 12 months (11).
- No postal address is published anywhere, and the homepage figures (60% cost reduction, 35% throughput) are given without methodology or source - treat them as vendor claims, not measured results.
Setup & Integrations
Technical difficulty
Moderate, and front-loaded on access rather than on engineering. Onboarding is AI-assisted - documentation ingestion, automatic system mapping, continuous learning - and pre-built connectors for CRM, ERP and databases mean no development work for standard cases. Fermi claims teams are operational in hours, and within days rather than weeks in its FAQ, with its own team available throughout. The real prerequisites are administrative: database credentials, Google OAuth authorization and Azure AD admin consent for Microsoft 365. A documented developer route exists (API key, service actor, token, Python and TypeScript SDKs), and the demo organization allows evaluation with nothing to install.
Deployment
Integrations
Supported languages
Behind Fermi Dev
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
How long does implementation take?
Does it work with the systems we already run?
How much does Fermi Dev cost?
Can I try it without signing up?
Is our data secure?
Will our data be used to train AI models?
How long is data kept?
Is there an API?
What size of organization is it built for?
How is it different from an AI chatbot?
Should you pick Fermi Dev?
Fermi Dev is technically well documented and unusually explicit about how it handles data. The API documentation is public, SDKs ship for Python and TypeScript, a pharmaceutical demo organization can be explored without signing up, and the security posture is named down to specifics: SOC 2 Type II, AES-256, SSO/SAML, role-based access control, audit logs that cover the actions of agents. The Simulate, Approve, Apply loop is the product's most distinctive idea - it answers head-on the obvious objection to letting agents act inside production systems. On governance, naming subprocessors in prose, keeping Microsoft 365 connectors strictly read-only and committing to deletion within 30 days is more than most vendors of this size put in writing. The main weakness is commercial opacity: no price, no plan, no rate card, no postal address. The Terms even point to a pricing page the site does not have. Budgeting means going through the sales team first. Several reservations should be settled before signing. The company is young - the domain was registered on 31/12/2024 - and two of the three demo organizations are still in progress. The homepage figures of 60% lower operating costs and 35% higher throughput are vendor claims published with no methodology and no source. And for a product aimed at regulated industries, there is no customer DPA, no documented training opt-out, no Article 27 EU representative and no explicit GDPR compliance claim, while section 4.2 of the privacy policy mentions learning from patterns in your organization's data. None of this is disqualifying, but all of it belongs on the list of questions to put to the vendor before a European deployment.
- Choosing a selection results in a full page refresh.
- Opens in a new window.