Transparent MCP proxy for agent observability. Logs every tool call, exposes trace history, reduces token overhead.
Transparent MCP proxy for agent observability.
Observer sits between your MCP client (Claude Desktop, Cline, Goose, Codex, etc.) and the actual MCP server. It logs every tool call to SQLite and exposes trace history through additional MCP tools, so you can see exactly what your agent is doing.
Across 1665 discussions in 8 major AI agent communities (MCP, Codex, Cline, Goose, Anthropic SDK, OpenAI Python), the 1 recurring problem is agent observability. Nobody can see what their agent is actually doing. There is no standard way to trace agent actions, tool calls, or decisions.
- Transparent proxy - Drop-in replacement for any MCP server. Your client does not know it is talking to a proxy. - Tool call logging - Every tool call is logged with input, output, duration, error status, and token estimate. - Trace tools - Exposes 4 additional MCP tools for querying history: - trace.history - List recent tool calls - trace.stats - Usage statistics with per-tool breakdown - trace.search - Search through tool call history - trace.replay - Replay a previous tool call by ID - Tool filtering - Hide tools from the client to reduce token overhead (addresses tool bloat, the 2 problem found in the same research). - SQLite storage - All data stored locally in SQLite. No external dependencies. - Zero config - One binary, one env var to point at your upstream server.
Or with Cline / Goose / any MCP client - just replace the server command with observer and set OBSERVERTARGET to the original command.
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/valtors--observer.html)
🛡️ Ultra-fast local security firewall, secret sanitizer, and context token compressor for Claude Code, Cursor, and MCP AI agents.
menot-you/n-memoryMemory your agent can trust — every answer carries its source; when it doesn't know, it says so. Hermetic, local, MCP.
enowdev/succubusCross-agent coordination for AI coding agents. One daemon, one database — so multiple agents in one repo can see each other's plan, tasks, and file claims.
Evermemoai/evermemoA tiny, universal memory engine for humans and AI agents — CLI, HTTP API, MCP server, and shared memory hub in one small Go binary
Smitner-Studio/facetRender agent-written Markdown as a live local page, and send the human's click back to the agent as typed data. Human-in-the-loop approval gates for coding agents.
Kota-Karthik/secure-actionsMCP server for secure AI agent authentication — lets Claude, Claude Code, and other LLM agents call APIs without exposing raw tokens or credentials to the model
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.