Melaya x Flowlines. Every MCP tool call observed. No screen, prompt or payload exported.
Privacy-first MCP observability: Melaya instruments its MCP server with Flowlines to see how every tool call performs without exporting a screen or prompt. Spans carry tool health, a pseudonymous account ID and explicit success or failure, never tool arguments, results, names or emails.
A stronger product on both sides
What this means for Flowlines
Teams connecting Claude, ChatGPT or Cursor to Melaya get an MCP server watched for silent failures and release regressions, while their phone screens, browser pages, prompts and business data never leave Melaya.
What this means for Melaya
Flowlines shows Melaya how 83 MCP tools behave across clients, scopes and releases, and the integration proves that useful observability needs no payload export: a pattern Melaya can extend to every agent surface.
A partnership, not a listing
What Melaya brings
- A governed remote MCP server with 8,277 scoped tools across 8 permission domains, OAuth 2.1 with PKCE, and reach into real Android apps, browser tabs, pipelines and connectors.
- A metadata-only span contract: a verified pseudonymous account ID, explicit success or failure, bounded error categories, and no tool arguments, results, names or emails.
- A privacy boundary enforced in code and tests, switched off until explicitly enabled and fail-open, so observability can never change what a tool returns.
What Flowlines brings
- Behavioral observability built for AI agents: sessions, users and cohorts, issues, releases and custom signals in one place.
- MCP-native ingestion over OpenTelemetry, alongside Claude Code, Codex, LangSmith and Langfuse as sources.
- A focus on the agent failures no one reports, so teams see where agents break and know when to act.
What we ship together
Compare MCP tool success rates and error categories before and after every Melaya server release, and catch a regression before users report it.
Find tools that fail repeatedly for one MCP client or permission scope, even when nobody files a bug, and fix the right one first.
See which clients and scopes drive real usage and which tools are never called, then shape the tool surface around evidence.
Build with Melaya and Flowlines
Talk to us about bringing this partnership into your stack, or read the announcement for the full story.
