Ambient intelligence for engineers: Why we made Suki's platform native to AI development tools with MCP

Suki's core product principle has always been meeting clinicians where they work. We don't ask a clinician to change how they practice medicine; ambient intelligence fits into the exam room as it already exists. When we looked at how developers building with Suki actually work in 2026, we realized the same principle should apply to Suki for Partners, and that our developer platform wasn't fully aligned with that principle

Developers building with Suki increasingly work AI-first. Their default motion is to describe intent to a coding agent: in Cursor, VS Code, or Claude Code, and review the output. Documentation that lives behind a docs site, a PDF, or a discovery call effectively doesn't exist in that workflow. So we made Suki's platform native to it with MCP support.

What we built

Suki's developer documentation is now available as a native MCP server. What this means for Suki’s partners is that any of their engineers working in an AI-assisted editor can add https://developer.suki.ai/mcp as a connector, and from that moment their coding assistant has direct, searchable access to Suki's full documentation and API reference. Alongside it, partners can install the Suki developer docs as a skill (npx skills add https://developer.suki.ai), giving their AI tooling durable context on our APIs, SDKs, and integration patterns.

The practical effect is subtle but profound. An engineer building against our Ambient APIs doesn't alt-tab to a docs site, skim for the right authentication flow, and translate it into code. They ask a question in their editor — "how do I authenticate and stream audio to a Suki session and retrieve the note?" — and the assistant queries our reference directly, pulling the correct WebSocket message formats, JWT requirements, and webhook configuration into its answer. The documentation stops being a destination and becomes ambient context.

How it works

The MCP server exposes two tools, with the singular goal of simplicity.

“Search” queries across our entire documentation set, including guides, API reference, SDK docs, and returns relevant snippets with links. "Get page" retrieves the full content from Suki’s documentation. AI applications chain these naturally: search to discover, get page to go deep. The search tool supports version and language filters, so an assistant working against a specific API version pulls only the reference that matches.

Setup takes under a minute in any major tool. Cursor and VS Code get one-click installs; Claude Code, Codex, and Gemini CLI get a single copy-paste command; anything else takes the raw endpoint URL. Both stdio and native HTTP transports are supported. There's no form, no credential exchange, no waiting on a partner enablement call to start building.

For AI tools that need broader context than tool calls provide, we also publish the standard tiers: llms.txt for a full documentation index, llms-full.txt for the complete corpus, and skill.md for teams building agents that integrate Suki's APIs directly.

Why it matters for the platform

Suki's platform thesis is that every piece of healthcare software will eventually have Ambient Clinical Intelligence built in, and the deciding question for those companies is build versus partner. Our answer to that question has always rested on integration velocity: partnerships that go live in weeks, not quarters, across our Platform products. Partner-proven APIs and SDKs are core to that speed, and MCP integration is now a central pillar enabling engineers to deploy faster without friction.

The tax on getting started with any platform is the integration work, and that tax is increasingly paid by engineers working through coding agents. When the agent has direct access to accurate reference material, integration questions get answered before they're ever asked: in the editor, at the moment of writing the code. When Suki meets with new partners, we can focus on optimizing workflow design, instead of getting stuck on documentation retrieval. The gap between a signed agreement and the first clinical note in production gets smaller, and the benefits of partnership are realized quickly.

It's also the first step in a longer arc. Today, a partner's coding assistant can read Suki. The direction we're building toward is a platform where partners compose ambient clinical workflows with the same fluency, where the integration itself becomes something you describe, not something you build. Readable documentation is the foundation that makes everything after it trustworthy.

The platforms that win in healthcare AI won't just have the best models. They'll be the easiest to build with for the clinician in the exam room, and now for the engineer in the editor.

Ready to build? Get in touch with us here.