Flowlines — Claude Code MCP server: session analysis and observability
Track how people use your MCP servers, which sessions fail and which intents changed after a release. Flowlines connects to your workspace and records telemetry from instrumented MCP servers.
MIT updated 2026-10-29
claude mcp add --transport http flowlines https://api.flowlines.ai/mcpWhat Flowlines does
Flowlines is an observability platform for MCP servers. The MCP server connects your client (Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI or others) to a Flowlines workspace at https://api.flowlines.ai/mcp. From there you see how tools are called, what results return, and where sessions break.
Four built-in skills handle different jobs: weekly reviews of what changed, before-and-after release comparisons, investigation of a failing session from a signal or complaint, and user cohort building by rule. The flowlines-doctor skill diagnoses connection issues and incomplete telemetry from MCP servers, LangSmith, Langfuse, and OTLP apps.
Start by running onboarding — it returns an instrumentation plan. Your API key is created once on the Flowlines setup page and shown only there; never paste it into chat. An instrumented MCP server ships tool arguments, client-visible results, and user metadata. You choose what to record on the initialization page.
What it helps with
- see which MCP server tools fail and where sessions break
- compare MCP server outcomes and costs before and after release
- trace a failing session from a signal down to the broken turn
- segment users by rule and measure the cohort against baseline
Example prompts for the agent
- Get an instrumentation plan:
/flowlines:onboarding - See what changed this week:
/flowlines:flowlines-weekly-review - Investigate why a session failed:
/flowlines:flowlines-investigate-session
How to install Flowlines
Platform CLI: remote server
claude mcp add --transport http flowlines https://api.flowlines.ai/mcpPlatform CLI: remote server
codex mcp add flowlines --url https://api.flowlines.ai/mcpConfig file
{
"mcpServers": {
"flowlines": {
"url": "https://api.flowlines.ai/mcp"
}
}
}Config file
{
"servers": {
"flowlines": {
"type": "http",
"url": "https://api.flowlines.ai/mcp"
}
}
}Platform CLI: remote server
gemini mcp add -t http flowlines https://api.flowlines.ai/mcpStep by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Flowlines
How do I install Flowlines in Claude Code?
Run claude mcp add --transport http flowlines https://api.flowlines.ai/mcp, then run /mcp and sign in to flowlines. Skills are available as /flowlines:flowlines-weekly-review or /flowlines:flowlines-doctor.
What does Flowlines do?
It is an MCP server for observing other MCP servers. It shows which tool calls fail, how an MCP server changed after release, and where sessions break. Built-in skills handle weekly reviews, release comparisons, and session investigation.
What data goes to Flowlines?
An instrumented MCP server exports tool arguments, client-visible results, and user metadata. This data can contain personal information, customer data, or source code. You choose what to record on the setup page; your API key is shown once and must never go into chat.