Source: Atomicwork Official Website Snapshot (Apr 2026)
Official-source follow-up to comprehensive-2026-04. This pass verified live atomicwork.com pages, Atomicwork’s Trust Center, TechCrunch funding coverage, and Microsoft Customer Stories.
What It Covers
- Current homepage positioning: “AI Agents that work for your IT team / employees / business” and 50%+ IT, HR, Finance, and Workplace request resolution from day one. ^[extracted]
- Current pricing page: Professional $90 / employee / year, Business and Enterprise custom, 4-week deployment promise, ROI guarantee, and 90-day ROI success plan. ^[extracted]
- Trust posture: SOC 2, ISO 27001, GDPR, CPRA, CSA STAR, HIPAA, ISO 27017, ISO 27018, ISO 27701, and ISO 42001 v2023 listed as compliant in the Trust Center. ^[extracted]
- Product architecture: Atom remains a multimodal Universal AI Agent across browser, Slack, Microsoft Teams, email, and business apps using chat, voice, and vision AI. ^[extracted]
- Microsoft GTM: Atomicwork is Microsoft Marketplace available, co-sell eligible, MACC-applicable, and explicitly Microsoft-first across Teams, Outlook, Edge, Entra, Intune, SharePoint, GitHub, Power BI, Sentinel, and Azure services. ^[extracted]
- MCP posture: Atomicwork does not appear to expose a public MCP server like Serval, but its workflow engine now publicly claims to use Claude Agent SDK, Claude Code, and MCP tools for tenant-aware workflow generation. ^[extracted]
Key Claims
Pricing and deployment
- Professional remains listed at $90 / employee / year with fair-usage access for Atom AI. Business and Enterprise are custom-priced. ^[extracted]
- The prior 4-6 week deployment claim should now be read as: pricing page promise = 4 weeks, while customer case studies still show 6-week incumbent replacement for Ammex and Pepper Money. ^[extracted]
- The “Atomicwork promise” includes dedicated onboarding, tailored enablement, specialist support, and a 90-day ROI success plan. ^[extracted]
Product surface
- Atom is positioned less as a passive assistant and more as a Universal Agent that “listens, sees, and responds” across apps, with screen-context reading, image analysis, voice guidance, and context-aware browser help. ^[extracted]
- The AI Agents page names Diagnostics Agent, Knowledge Agent, Operations Agent, and Incident Agent; the homepage separately highlights IGA Agent and Onboarding Agent. ^[extracted]
- Agentic Enterprise Automation claims natural-language, drag-and-drop, and code workflow authoring, with generated branching logic, system actions, and custom code “in minutes.” ^[extracted]
- Workflow automation supports scripts, arbitrary API calls, scheduled workflows, AI classification, version controls, JSON transforms, error logs, retries, and rollback-oriented versioning. ^[extracted]
Security and compliance
- Atomicwork’s Trust Center lists 10 compliance frameworks: SOC 2, ISO 27001, GDPR, CPRA, CSA STAR, HIPAA, ISO 27017, ISO 27018, ISO 27701, and ISO 42001 v2023. ^[extracted]
- Security page adds Microsoft 365 certification and CASA for Google Apps as enterprise trust signals. ^[extracted]
- Customer data controls include per-tenant vector databases, customer control over indexed/modified/deleted data, no user data used for model training or tuning, no third-party sharing, PII redaction at ingestion, and PII masking in Atom conversations. ^[extracted]
- Platform page mentions region-specific hosting options, but no public self-hosted or hybrid deployment model was found. ^[extracted]
- TechCrunch reports Atomicwork can let enterprises own encryption keys or bring their own model endpoints; this is a stronger enterprise-control claim than the marketing pages alone disclose. ^[extracted]
MCP and workflow generation
- Atomicwork’s MCP thought-leadership post frames MCP as “Enterprise IT’s new TCP/IP moment” and says Atomicwork is leveraging MCP to connect enterprise tools while maintaining security and context awareness. ^[extracted]
- The Claude workflow-engine post goes further: Atomicwork says agents query live MCP tools to inspect triggers/actions, schemas, field types, and valid values, then construct typed execution plans with explicit dependencies and branching logic. ^[extracted]
- Atomicwork says it generates tenant-specific TypeScript SDK bundles from configured MCP servers. Each bundle can only call that tenant’s servers with that tenant’s credentials/endpoints, then runs in an isolated sandbox. ^[extracted]
- No evidence was found of a public Atomicwork MCP endpoint or external MCP server docs. The public developer site exists but rendered only a documentation shell during this pass. ^[ambiguous]
Strategic Implications
- Atomicwork has moved from “MCP narrative only” to MCP-backed internal workflow generation. That does not equal Serval-style public MCP distribution, but it closes part of the architecture gap around tool/schema discovery.
- The 4-week deployment promise is now more aggressive than the existing 4-6 week framing, though case-study evidence still centers on 6-week replacements.
- Compliance depth is stronger than previously captured: the important additions are ISO 27017, ISO 27018, ISO 27701, CSA STAR, and CPRA.
- Atomicwork is still not showing a public self-host / hybrid story. The enterprise-control story is instead region-specific hosting, per-tenant data isolation, customer-owned keys, and bring-your-own model endpoints.
Limitations
- Most claims are vendor-authored and should be treated as marketing evidence unless independently corroborated.
- Public MCP endpoint absence is inferred from search and available docs, not from an explicit Atomicwork denial.
- The developer documentation shell did not expose endpoint details via WebFetch; browser-based inspection may reveal more.