Content Intelligence MCP Server
The Content Intelligence MCP server (CIN MCP) connects AI assistants, such as Claude, to Hyland Content Intelligence, allowing your users to search, read, and act on their content in plain language, all under their own Hyland account and permissions.
What is CIN MCP?
CIN MCP is a remote server that uses the Model Context Protocol (MCP), the open standard AI assistants use to call outside tools. An assistant connected to it can find and read Content Lake documents, ask Knowledge Discovery agents questions, build and run Agent Builder agents, and inspect agent traces. One connection covers all of these products.
Every user signs in with their own Hyland account. Each tool call runs with that user's environment, subscriptions, and product roles, allowing the assistant to see and do only what the user could do in the products themselves.
Key Features
CIN MCP offers the following features:
One Connection, Many Products
90 tools are available across Content Lake, Enterprise Search, Knowledge Discovery, Agent Builder, and Trace Observability. The assistant can combine them in one conversation: find documents in Content Lake, read them, and then ask a Discovery agent a follow-up.
Your Permissions, Not a Service Account
OAuth sign-in is available with each user's Hyland account. The server exchanges the user's token for a product-specific token on every call, and each product applies its own roles. There are no shared service accounts or stored passwords.
Safe By Design
CIN MCP is safe by design for the following reasons:
- Every tool declares whether it only reads or changes something, so assistants and administrators can require approval for changes.
- Error messages carry a correlation ID for support. Also, they don't expose secrets, tokens. or document text.
- Agent Builder guardrails can be set on an agent or applied to a single request.
Works Where Your Users Already Work
The Claude desktop app and Claude Cowork work through an organization connector, Claude Code, OpenAI Codex, and Agent Builder agents, which can call CIN MCP tools themselves. Supported clients are desktop only; web clients in a browser, such as claude.ai, aren't supported.
Architecture Overview
AI client (Claude desktop, Cowork, Hyland IAM
Claude Code, Codex, Agent Builder) (sign-in, token exchange)
│ MCP over HTTPS ▲
│ Bearer token (user) │ exchange per product
▼ │
┌──────────────────────────────────────────────┐
│ CIN MCP server │
└──────────────────────────────────────────────┘
│ product token (user's own roles)
▼
Content Lake · Enterprise Search · Knowledge Discovery · Agent Builder · Trace Observability
Component Overview
CIN MCP consists of the following components:
| Component | Role |
|---|---|
| AI client | Runs the conversation, decides which tools to call, and shows results to the user. |
| Hyland IAM | Signs the user in (OAuth authorization code with Proof Key for Code Exchange (PKCE)) and exchanges the user's token for a token per product. |
| CIN MCP server | Exposes the tools, validates input, calls the products with the exchanged token, and returns safe, structured results. |
| Content Intelligence products | Do the work and enforce the user's subscription and roles. |
Use Cases
Some example use cases for CIN MCP include the following:
Enterprise Knowledge In the Assistant
- Ask Knowledge Discovery agents questions and get cited answers.
- Search Content Lake by name, metadata, or content, and read the documents directly.
Agent Building and Debugging
Solution builders create, configure, and test Agent Builder agents from a conversation, and then use Trace Observability to see where an answer came from and where the time went.
Knowledge Graph Work
With the Enterprise Context Engine, you can do the following:
- Inspect ontologies and knowledge graphs.
- Run read-only graph queries.
- Check graph health.
- Manage graph rebuilds.
Quick Navigation
| Page | What's In It |
|---|---|
| Getting Started | Prerequisites, the setup sequence, and a first conversation |
| Subscription and Roles | Subscriptions, users, groups and roles in the Content Innovation Cloud Administration Portal |
| Authentication | How sign-in, scopes and token exchange work |
| Tutorials | Connecting the Claude desktop app, Claude Code, Codex, and Agent Builder |
| Tool Reference | All tools by product, and whether each reads or changes data |
| Error Reference | Errors by category: what they mean and what to do |
Tool Overview
| Area | Tools | What It's For |
|---|---|---|
| Agent Builder | 32 | Build, configure, run, and monitor agents, guardrails, async runs, and registered MCP servers. |
| Content Lake | 34 | Find, browse, read, and query documents, metadata, ontologies, and knowledge graphs. |
| Enterprise Search | 1 | Perform one hybrid search across Content Lake passages, full text, and metadata. |
| Knowledge Discovery | 5 | Ask multi-turn questions to Discovery agents with citations. |
| Trace Observability | 18 | Observe traces, sessions, latency, evaluations, and run metrics. |
The Tool Reference lists every tool.
Response Quality
Answers are only as good as the content, the agents, and the tools behind them.
- Tools return what the products return. Graph and Discovery answers depend on how the content was ingested, enriched, and modeled.
- The assistant chooses which tools to call. Clear questions that name the product or agent (for example, "Ask the HR Discovery agent") get better results.
- For decisions, read the cited source documents; don't rely on a summary alone.