MCP Integration
Stage: platform (MCP server) · What it does: expose your catalog, data, and models to AI agents through the Model Context Protocol (MCP) — the open standard Claude/Copilot-class agents use to reach tools and data.
Why it matters
- Your data, agent-ready. Instead of writing bespoke glue for every agent, one MCP server publishes your governed data sources as agent-callable tools.
- Governed. What an agent can see is what your RBAC says it can see — no "agent with a bucket key" sprawl.
- Open standard. Any MCP client works.
How to use
- Deploy/start the platform's MCP server for your tenant.
- Connect an MCP client (Claude, Copilot, your own agent runtime) to the server.
- Grant the agent access to the data sources you choose (scoped by role/permission).
- The agent can now query/call your data and your models through the same governed surface.
Real-world examples
🧑💼 SaaS — support agent with the actual product data
A support team wires their agent (via MCP) to the product's support catalog and documentation. The agent answers with their content, not a generic model's guess — and the access is scoped, so it can't reach customer PII.
🏦 Financial services — analyst copilot over governed sources
Analysts use a copilot that reaches the internal data catalog through MCP. The copilot can only touch datasets the analyst's role permits — the agent inherits RBAC, so "ask the copilot" never becomes "the copilot has a master key."
🛒 Retail — agentic merchandising
A merchandising agent reads current inventory/sales via MCP and drafts restock notes using the team's fine-tuned model via the inference gateway — two governed surfaces, one workflow.
Gotchas
- MCP access is permission-scoped — grant the least access the agent needs.
- Agent tool calls are auditable like other platform actions.
- Keep the MCP server on the same hosting you choose for everything else — hosted, BYO AWS, or self-hosted on AWS.