Back to vendors
U

Unqork

Also known as: UnqorkAI, Unqork Build Agent, unqork.com

Visit site
Entry priceQuote only, after a demoFull pricing detail

Unqork is an enterprise application platform where the UnqorkAI Build Agent turns prompts, designs and requirements documents into governed applications. The agent proposes a plan for approval, builds from reusable components rather than writing new code, writes tests and documents the result, and keeps a shared workspace memory of what it learns. Customers run the apps on Unqork's single tenant cloud and can add their own conversational agents to them with the models they choose.

Unqork, led by founder and CEO Gary Hoberman, sells an enterprise application platform to large financial services, insurance, healthcare and government organizations. Applications are assembled from reusable components on one runtime engine rather than written as code, so an update to a component rolls out to every app that uses it. Customers shown include Goldman Sachs, BlackRock, Marsh, Hitachi, Principal Financial Group, Broadridge, Allianz Commercial, AXIS Capital and the Federal Home Loan Bank of New York, and Voya Financial chose UnqorkAI for its retirement experiences.

UnqorkAI adds a Build Agent inside the Module Builder. A creator describes what they want, or hands over designs, specifications, requirements documents and data sources connected through MCP, and the agent proposes an implementation plan that must be approved before it builds.

It then configures the app from components, connects the logic, writes test plans and tests, and generates documentation for new and existing apps, reusing components instead of recreating them to keep token spend down. A reasoning panel shows its steps, effort levels shift during a task, and task history stays for the life of the environment.

The agent keeps Context Documents for each workspace, writing and consolidating entries as it works so every creator and every later task starts from the same knowledge.

Governance follows the platform's own rules. The Build Agent works only in non production environments, inherits the signed in creator's role based access, cannot reach UAT or production, and asks for approval on every change by default. Guardrail alerts go to a security information and event management system for a year, and audit logs record platform activity.

Inside the finished apps, Agentic Smart Components such as a chat agent run on models the customer picks from Google Gemini, Amazon Bedrock, Microsoft Azure or Anthropic, while the Build Agent itself runs on Google's Gemini platform. A beta MCP server publishes API modules so outside agents can call them, and the marketplace adds prebuilt solutions such as an AI Underwriting Workbench from Unqork and accelerators from EY, Quantiphi and Slalom.

Unqork sells through demos, and UnqorkAI requires an AI capabilities agreement. The platform holds SOC 1, SOC 2 Type II and ISO/IEC 27001 reports, appears on the FedRAMP Marketplace, supports PIV and CAC authentication for government users and runs single tenant instances with separate non production environments, processing data in the United States. It suits regulated enterprises that want AI to speed up building governed business applications without unreviewed code. Teams that want generated source code to take elsewhere, or hosting outside the United States, will find a closed platform.

Vendor details

Canonical URL

https://unqork.com/

Category

Coding agent

Subcategory

AI build agent for governed enterprise applications

Company status

independent

Use cases & customers

Primary use cases

Build governed enterprise applications from requirements with an AI agentModify and reuse components across many applications at onceAdd conversational agents to applications with customer chosen models

Target customers

Financial services and insurance enterprisesHealthcare organizationsGovernment agencies

Deployment options

Single tenant Unqork cloud with separate non production environmentsFedRAMP listed government offering

Integrations

Third party APIs, SFTP and S3 batch transfer, BI connectors, MCP connectors for build context, Agentic Smart Components on Google Gemini, Amazon Bedrock, Microsoft Azure or Anthropic, and an MCP server that publishes API modules to outside agents.

In practice

An insurer hands the Build Agent a requirements document for a claims intake app, approves its plan and reviews the tests it writes.

A bank updates a shared identity verification component once, and every application that uses it picks up the change.

A retirement services team adds a chat agent on Anthropic models to a participant portal built on Unqork.

Agentic Index coverage score

10.5 / 14 capabilities · 75%

Integrations & Tool Calling Full

Agentic Smart Components inside finished applications call approved MCP tools as they work, and applications reach third party systems through APIs, SFTP and S3 batch transfers and BI connectors. While building, the agent can draw on data sources connected through MCP connectors, which are in beta. The Build Agent itself acts inside the Module Builder, configuring components and logic in the customer's own workspace.

Sourceunqork.com/product-overviewread 2026-11-20

Workflow Orchestration Full

The Build Agent works through planned stages: it turns inputs into an implementation plan, builds components and connects logic, writes and runs tests, then documents the app, with each task broken into steps it reasons through in a visible panel. The applications it produces run multistep business processes on Unqork's workflow engine, which customers have used to blend automation, AI and workflow in one place.

Sourcedocs.unqork.io/unqorkai/docs/unqorkairead 2026-11-20

Knowledge Grounding & RAG Partial

Each task draws on the material a creator supplies, such as prompts, designs, specifications and requirements documents, along with data sources connected through MCP and the chat context for that task, which resets between tasks. The agent can answer questions about how a module is configured by reading it. No index or embeddings store over the customer's wider document estate sits behind those answers.

Sourcedocs.unqork.io/unqorkai/docs/unqorkairead 2026-11-20

Human Oversight & Guardrails Full

The Build Agent proposes an implementation plan that a creator must approve before it builds anything, and by default every change it makes needs the creator's approval too. It works only in non production environments, cannot reach UAT or production, and carries no privileges beyond the signed in creator's role. AI guardrails flag risky prompts and responses for review.

Sourcedocs.unqork.io/docs/data-privacy-and-security-for-unqorkairead 2026-11-20

Security, Identity & Governance Full

Role based access control applies to each promotable item type with separation of duties across roles, single sign on runs through OIDC providers such as Amazon Cognito, and government users can authenticate with PIV or CAC cards. Unqork holds SOC 1, SOC 2 Type II and ISO/IEC 27001 reports, shared through its trust portal, and appears on the FedRAMP Marketplace.

Sourceunqork.com/security-complianceread 2026-11-20

Observability & Auditability Full

Build Agent prompts and responses stay in task history for the life of the environment in the customer's own database, and a reasoning panel shows how the agent worked through each step. Platform audit logs record activity, guardrail logs are kept for 30 days, and prompts that set off a guardrail alert stay in the security event system for a year, alongside trace and span telemetry.

Sourcedocs.unqork.io/docs/data-privacy-and-security-for-unqorkairead 2026-11-20

Memory & State Persistence Full

Context Documents give each workspace a shared memory that the Build Agent reads at the start of every task and writes as it works, explaining each change in chat and consolidating entries as they grow. Entries last until someone deletes them, with earlier versions kept in history, workspace permissions decide who can view, edit or clear them, and an administrator can switch the feature off for an environment or workspace.

Sourcedocs.unqork.io/docs/context-documentsread 2026-11-20

Deployment & Data Residency Partial

Each customer gets a single tenant instance by default, with several non production environments in a VPC separate from production and data stored in MongoDB Atlas, and government customers can use the FedRAMP listed offering. All processing happens in the United States, so buyers can choose between commercial and government environments but not a region outside the US or their own infrastructure.

Sourceunqork.com/security-complianceread 2026-11-20

Prebuilt Agents / Templates / Packs Full

The marketplace offers prebuilt snippets, templates and integrations alongside complete solutions, among them Unqork's AI Underwriting Workbench, EY's Add Insurance Now and Global Resource Tracker, Quantiphi's AiUP with Google and Slalom's Lightflow with AWS. Each solution covers its own business process, and Smart Components such as the Chat Agent arrive ready to configure with the Build Agent.

Sourceunqork.com/marketplace-solutionread 2026-11-20

Triggers & Channel Coverage Partial

The Build Agent starts when a creator asks it for something in the Module Builder, with prompts, designs, specifications or requirements documents as input. Agents built into applications respond through those applications' own screens and API modules, which the beta MCP server can also open to outside agents. Schedules, webhooks or system events that start the Build Agent without a creator have no place in the product.

Sourcedocs.unqork.io/unqorkai/docs/unqorkairead 2026-11-20

Model Flexibility & Routing Partial

Agentic Smart Components in finished apps run on models the customer picks from Google's Gemini platform, Amazon Bedrock, Microsoft Azure or Anthropic, configured with instructions, temperature, thinking level and token limits per LLM configuration. The Build Agent, which is how Unqork builds applications, runs on Google's Gemini platform with Unqork choosing the models, so the choice stops short of the builder itself.

Sourcedocs.unqork.io/docs/data-privacy-and-security-for-unqorkairead 2026-11-20

APIs / SDKs / MCP Extensibility Full

API modules expose application logic as callable endpoints secured by API credentials tied to Express roles, and a beta MCP server publishes those modules as tools that outside agents can list and call, with a guide for building such an agent in Node.js. Custom components, functions and a plug in component let developers extend what the platform and the Build Agent can assemble.

Sourcedocs.unqork.io/docs/mcp-guidesread 2026-11-20

Testing, Debugging & Optimization Partial

As part of each build, the agent generates a test plan, refines edge cases and implements tests for the application, and creators review its output because suggestions can be wrong. Those tests check the app the customer owns rather than the agent's own behavior. Guardrail alerts flag risky prompts, but no benchmark or scored evaluation measures the quality of the Build Agent's work over time.

Sourceunqork.com/product-overviewread 2026-11-20

Browser / Computer-use Not documented

The Build Agent works inside Unqork's own Module Builder, configuring components through the platform rather than operating outside software, and finished applications connect to other systems through APIs, MCP tools, file transfers and BI connectors. Neither the builder nor the agents inside apps drive a browser, a desktop session or a remote computer.

Sourceunqork.com/product-overviewread 2026-11-20

The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded

Pricing

Quote only, after a demo

Enterprise platform subscription plus an AI capabilities agreement

What is public

Unqork sells its application platform through a demo and a quote. UnqorkAI, including the Build Agent, requires an active AI capabilities agreement, and customers on Unqork 9.0 upgrade through their Unqork representative.

Billing mechanics

The enterprise agreement sets the environments, applications and AI capabilities covered. Agentic Smart Components inside apps run under the customer's own agreement with the model provider it chooses.

Sales call required

Yes, required for paid access

Free / trial

Demo

Commercial notes

Partners such as EY, Quantiphi and Slalom build marketplace solutions on Unqork, and Voya Financial chose UnqorkAI for its retirement experiences.

Missing data

Unqork sets platform and AI fees in each enterprise agreement after a sales conversation.

Agentic Index verified 2026-11-20

Contact us

Found a vendor we missed? Have feedback on the index? We'd love to hear from you.