Your API was built for another audience.
The API may expose low-level operations without enough context for an agent to use them well. You need a smaller, intentional set of tools and useful descriptions.
Connect / Model Context Protocol
Make selected instrument capabilities usable by an AI agent.
We develop MCP servers around instrument APIs and SDKs, giving compatible agent clients a defined set of tools to discover and use. The work starts with the instrument’s actual behavior: operations, state, completion and errors.
connect / Conceptual model
The customer problem
Where the next capability gets blocked.
Instrument operations take time.
A command can be accepted before an operation finishes. The integration needs to report state, errors and uncertain outcomes in a form the client can handle.
Access needs clear boundaries.
Discovering a capability should not grant unrestricted execution. Tool access, credentials and required approvals need to fit your application.
What we build
A useful scope.
A reviewable result.
We scope the server around one instrument family, selected operations and a target client. The result is evaluated through representative tool calls and workflow scenarios.
- 01
A curated tool interface
Translate selected API, SDK or SiLA 2 capabilities into described tools with input schemas, documented behavior and useful results.
- 02
Instrument-aware operation handling
Implement the agreed lifecycle for commands, long-running operations, errors and cancellation. Define how the client learns what completed and what remains uncertain.
- 03
Client integration and access controls
Test the server with the selected MCP client and transport. Agree authentication, credentials, exposed operations and approval responsibilities across the application.
- 04
Reference scenarios and documentation
Provide representative calls, integration guidance and acceptance tests. Add modeled workflow checks where the use case requires them.
Relevant engineering
Our engineering notes explain the practical issues in wrapping vendor SDKs for instrument-facing agent tools.
Before we start
A few practical questions.
What does an MCP server do for an instrument?
It exposes selected instrument capabilities as tools that a compatible AI client can discover and call. An adapter connects those tools to the underlying instrument API or SDK.
Which clients will it work with?
We agree the target MCP client, protocol features and transport, then test that combination. Compatibility with another client should be verified against its supported features and your workflow.
Does MCP replace SiLA 2 or a vendor SDK?
No. Those interfaces can remain underneath the MCP server. The appropriate architecture depends on the existing software and how agents and other lab systems need to access it.
How do we start?
Bring the available API or SDK documentation, the operations you want an agent to use and a target client. We assess readiness and define an initial integration milestone.
Your next project
What’s next for your instrument?
Tell us what you want to connect, simplify or automate. Let’s define the first useful step together.