OpenAI

Back to Providers

Use this page to configure an OpenAI connection for the current v0.6.x Early Access lane.

Scope and release-lane context

This guide is for the current v0.6.x Early Access lane of mcp synapse.

Boundary conditions for this lane:

BYOK/local-only
not a hosted proxy service
non-streaming for this lane
no hidden retry, backoff, or silent fallback

This page covers a basic OpenAI connection flow for local runtime usage, preflight, and a minimal validation check.

Prerequisites

A local mcp synapse runtime environment
An MCP-capable IDE or client
An OpenAI API key that you control
Access to at least one OpenAI model identifier you are allowed to use
Outbound network access from your local machine to OpenAI endpoints

Core concepts

The provider identifier for this guide is openai.
Authentication is key-based.
Requests in this lane are non-streaming.
Connection preflight should pass before IDE-side usage.
Provider credentials stay under your local control.

Setup steps

Start your local mcp synapse runtime.
Open the Connections surface in mcp synapse.
Create a new OpenAI connection.
Provide your key in the runtime-supported local credential path or method.
Set a valid OpenAI model identifier.
Save the connection.

If your environment allows custom endpoint configuration, keep it explicit and valid for OpenAI API usage.

Connection settings

Use only settings that are clearly supported by your current build and UI.

Minimum operational settings:

Provider: openai
Authentication: local key under BYOK control
Model: a valid model identifier you can access

Avoid adding optional tuning fields unless you have a confirmed operational need.

Preflight and IDE integration

Run connection preflight in mcp synapse.
Confirm preflight reports success.
In your MCP-capable IDE, select the OpenAI-backed route or connection.
Send a simple test prompt.
Confirm a normal text response is returned.

Use one minimal prompt first, then expand to your real workload.

Troubleshooting

Auth failure

Verify the API key is present, readable by the runtime path, and not empty.
Rotate or re-enter the key if needed.

Model access failure

Confirm the model identifier is valid and accessible for your account.
Try a known-good model identifier in the same account.

Endpoint or config issues

Check endpoint or base URL formatting if customized.
Ensure local network egress to OpenAI is not blocked.

Preflight fails

Resolve the explicit preflight error first.
Re-run preflight before IDE testing.

Security and key handling

Keep OpenAI keys local and out of source control.
Do not commit secrets to repository files.
Prefer least-privilege operational key handling where possible.
Treat logs and screenshots as potentially sensitive.

Operational boundary reminder:

mcp synapse in this lane is BYOK/local-only.
It is not a hosted proxy service.

Minimal checklist

Local mcp synapse runtime is running
OpenAI key is configured locally
Provider is set to openai
Model identifier is valid
Preflight passes
IDE test prompt returns a normal response
No secrets are committed