The answer
Open Use with and choose Claude Code or Codex. This connects your assistant to the Flowlines MCP. Connecting alone does not instrument your server or enable coding-tool monitoring. Sign in, then choose the setup you need.
For an application agent, book a setup meeting and bring the provider already used by the application. For OpenTelemetry, send OTLP/JSON or protobuf traces to https://api.flowlines.ai/v1/traces with the key in thex-flowlines-api-key header.
If traces already live in Langfuse or LangSmith, use read-only API polling. Both providers also support a push path based on OpenTelemetry export. A public LangSmith trace share can be imported once for a focused investigation.
MCP servers use a different path. The Flowlines repository skill reviews the server and the proposed export boundary, then guides an OpenTelemetry-compatible integration after approval.
For coding-tool monitoring, use the official Flowlines plugin and explicitly enable the relevant integration. Review the prompts, responses, tool activity, and usage details it will export. If a newly installed plugin is unavailable, start a new Claude Code session or Codex task. If you need a namespace API key, the setup guides you to create one in Flowlines.
How it works
Three sources, three deliberate paths
Coding-agent CLIs use the global plugin. Existing application traces connect through standard transports or supported providers. MCP servers add compatible instrumentation. Each path keeps the data boundary explicit.
Last reviewed 2026-09-25