> ## Documentation Index > Fetch the complete documentation index at: https://docs.prisme.ai/llms.txt > Use this file to discover all available pages before exploring further. # Custom portals & apps > Replace scattered internal tools with one custom agentic portal: a React front-end, agents, and workflows connected to your systems, governed, on SaaS or self-hosted. Enterprises rarely lack tools. They have too many: a CRM here, a ticketing tool there, spreadsheets, a legacy app nobody dares to touch, and no single place where it all comes together. The highest-value pattern on Prisme.ai is to build **one custom portal** that unifies them: a React front-end, agents, and workflows, connected to your systems and governed end to end. It is the same thing every Prisme.ai product is: a workspace built in [Builder](/products/ai-builder/overview). The products are the proof; your portal is built the same way. One pattern for custom portals: your systems and data (CRM, ITSM, databases, files, legacy apps) connect via connectors and MCP into a portal built in Builder (React front-end, agentic workflows, agents, Knowledges and Collections), delivered to your channels (web portal, Teams, Chat, embed), governed and traceable, deployed on SaaS or self-hosted. ## One pattern, many portals Same building blocks, different business domains. A sales and account workspace where agents draft, summarize, and update records, on top of your CRM. Ticket triage, resolution, and knowledge lookup for IT and internal support, connected to your service desk. Watch regulations, assess impact, generate documentation, and keep an audit trail for GDPR and the AI Act. Onboard, score, and monitor suppliers, with agents that read contracts and flag risks. One front-end over applications that do not talk to each other, orchestrated by workflows instead of manual copy-paste. Aggregate alerts, enrich with context, and drive response, on top of your security tools. Connect entities across your data and let agents navigate and answer over the graph. Short-lived portals for an event, a crisis cell, or a campaign: stand one up fast, retire it when done. ## Modernization, not rip-and-replace You do not have to rewrite your systems. A portal connects to them non-invasively through [connectors and MCP](/apps-store/overview), and adds the experience and the agents on top. It is a pragmatic path to **modernize legacy** and **rationalize software costs**: keep what works, wrap it in one governed experience, and retire the tools that were only there to fill a gap. ## How it is built Bring in CRM, ITSM, databases, files, and legacy APIs through connectors and MCP. No rebuild. Compose agents with your skills and knowledge, and agentic workflows for the business logic. A React front-end in [Builder](/products/ai-builder/overview), starting from the [starter SPA](https://github.com/prismeai/starter-spa) and versioned in Git, so the portal is real software, not a black box. Roles, audit, and quotas from [Governe](/products/ai-governance/overview); deliver to a web portal, Teams, Chat, or an embed. ## Built with your coding agents You do not have to build a portal by hand. Because the whole platform is reachable from **VS Code over [MCP](/products/ai-securechat/mcp-connections)**, you can drive it with the coding agents your teams already use, Claude, GitHub, and others, to scaffold pages from the [starter SPA](https://github.com/prismeai/starter-spa), write automations, and wire connectors. And because Prisme.ai runs on a clean, scalable Information System that **mutualizes the infrastructure**, what those agents generate lands on shared, governed foundations rather than a new stack per project. ## The economics: it amortizes Cost is where the pattern really pays off. The classic path gives every use case its own dedicated app, CI/CD pipeline, Ops, SIEM, observability, and access layer. That scales linearly: twice the use cases, roughly twice the cost. Total cost vs number of use cases. Classic delivery rises steeply and linearly, a dedicated app, CI/CD, Ops, SIEM, observability, and access per use case. Prisme.ai starts with a shared platform cost then rises gently, since each new use case adds a use case, not a stack. The lines cross at a break-even point, after which Prisme is cheaper. On Prisme.ai, the cost is **sublinear**. What you set up once is shared by every use case: * **Infrastructure** mutualized across all workspaces * **Governance and access** (SSO, roles, quotas, audit) once, for everyone * **Observability and SIEM** integration shared, not rebuilt per app * **Software licenses** reduced as one portal replaces several tools Each new use case is a fraction of the cost of the first. Past a handful of use cases the platform has amortized, and every additional one widens the gap. ## SaaS or self-hosted Start on **SaaS** to move fast and prove value. Move to **self-hosted** when criticality, data sovereignty, or regulation require it: the same portal, versioned in Git, runs on your own cloud or on-premises without a rewrite. See [Self-Hosting](/self-hosting/overview) and [the sovereign agentic layer](/get-started/sovereign-agentic-layer). ## Next steps Build the portal: pages, automations, imports Environments and the two-workspace Git model