In one line: MCPcat is closed-source and hosted; mcpeye is open-source (MIT) and self-hosted, and its hero feature — the Intent Gap Report — tells you not just what happened, but what to build next.
mcpeye vs MCPcat at a glance
| mcpeye | MCPcat | |
|---|---|---|
| License | Open source (MIT) | Closed source |
| Hosting | Self-hosted (docker compose up) | Hosted SaaS only |
| Your data leaves your stack | Never | Sent to their cloud |
| MCP session replay | ✅ | ✅ |
| Agent-goal analytics | ✅ | ✅ |
| Intent Gap Report (build-next roadmap) | ✅ | ⚠️ partial |
| Missing-capability capture | ✅ | ❌ |
| SDKs | TypeScript, Python, Ruby | — |
| Price | Free, forever | Paid |
Why teams choose the open-source option
Your data never leaves your stack
MCP tool calls contain your users’ literal requests and arguments — the most
sensitive product data you have. With mcpeye, Postgres, Redis, and your LLM key
are all yours. Nothing egresses to a third party.
A build-next roadmap, not just dashboards
The Intent Gap Report surfaces the asks your tools couldn’t deliver and the
capabilities agents wanted that no tool exists for yet — ranked, with the exact
sessions behind each one.
Open source you can audit and extend
Read the code, self-host it, fork it, send a PR. No black box deciding what your
agents’ data is worth.
One-line SDK, three languages
track(server, "id") in TypeScript, Python, or Ruby. The agent self-reports its
intent inline, so capture is near-free and no LLM runs on the hot path.Switching is one command
Get started in five minutes
Self-host mcpeye and read your first Intent Gap Report.