Spanly

Product Introduction

1. Definition: Spanly is a dedicated observability platform specifically engineered for Model Context Protocol (MCP) servers. It functions as a specialized monitoring layer, often deployed via a drop-in CLI wrapper or SDK, that captures and analyzes the full spectrum of MCP traffic—every tool call, prompt, resource read, and JSON-RPC packet.
2. Core Value Proposition: Spanly provides comprehensive MCP server observability to engineering teams shipping AI-powered features in production. It eliminates the blind spots of generic APM tools by offering deep, protocol-level insights into error rates, session traces, latency metrics, client analytics, and deployment impacts, enabling rapid debugging and performance optimization for AI agents.

Main Features

1. Protocol-Level MCP Monitoring: Spanly operates at the transport level (stdio or HTTP), capturing complete JSON-RPC packets without requiring code changes to the server application. This means every tools/call, prompts/get, and resources/read request and response, including full payloads, arguments, results, and error codes, is traced end-to-end.
2. Zero-Code, Multi-Language Instrumentation: The platform works with any language that implements the MCP specification. Engineers can wrap any existing MCP server binary with the Spanly CLI, drop in the TypeScript or Python SDK, or deploy as a Docker sidecar. This provides instant visibility with no schema maintenance, agent processes, or rebuilds required.
3. Comprehensive Performance Analytics: Spanly breaks down performance metrics by server, client, tool, prompt, and resource, offering percentile latencies (P50, P95, P99) and volume analytics. It helps teams pinpoint slow operations, track error rate trends, and identify performance regressions tied to specific client versions or deployments.
4. Advanced Session and Error Tracking: The system groups individual requests into interactive sessions, tracing the full arc of an interaction across multiple tool calls. Recurring errors are deduplicated and tracked by code, providing clear patterns for debugging issues like Invalid Params or Request Timeout.
5. Integrated Alerting and Status Boards: Spanly includes configurable alerts on metrics like error rate spikes, p99 latency thresholds, or traffic drops, with routing to email, Slack, and webhooks. It also supports always-on status boards for live monitoring of uptime, error counts, and open incidents.
6. Data Residency and Compliance: The platform offers strict data residency controls with ingest and storage endpoints in both the US and EU regions. This design ensures GDPR compliance for EU customers, with no cross-region data replication.

Problems Solved

1. Pain Point: Generic Application Performance Monitoring (APM) tools like Datadog or New Relic lack deep, out-of-the-box visibility into the MCP protocol. Their instrumentation is often limited to specific languages, requires manual code changes to capture spans, and doesn't provide the full JSON-RPC packet context needed to efficiently debug failing tool calls in production AI systems.
2. Target Audience: Spanly is built for SaaS engineering teams, DevOps, and platform engineers who are building, deploying, and maintaining production MCP servers. This includes teams using MCP clients such as Claude Code, Cursor, Copilot, and other AI agents.
3. Use Cases: It is essential for scenarios such as diagnosing why a specific AI tool call fails intermittently, monitoring the performance impact of a new MCP server deployment across client versions, identifying which MCP tools are most heavily used in production, and maintaining SLAs for AI-powered features that rely on MCP.

Unique Advantages

1. Differentiation: Unlike traditional APM add-ons that operate as a vendor SDK inside the server process, Spanly acts as a complementary, dedicated layer. It captures protocol-shaped detail that APMs miss, provides full payload visibility without hitting tracing quotas, and supports any language or transport method with zero code changes. It explicitly complements, rather than replaces, the existing Datadog, Sentry, or New Relic stack.
2. Key Innovation: The key innovation is its transport-level approach to observability. By intercepting the MCP traffic stream itself, Spanly guarantees universal compatibility and captures the complete request/response data in its native format. This allows for per-client, per-tenant analytics derived from headers or JWT claims, and a proxy mode that can monitor third-party servers the engineering team does not own.

Frequently Asked Questions (FAQ)

1. What is MCP observability and how does Spanly implement it? MCP observability refers to the monitoring, tracing, and analysis of traffic within the Model Context Protocol, which standardizes communication between AI clients and servers. Spanly implements this by capturing all JSON-RPC messages at the protocol level, providing full traces of tool calls, errors, latency, and usage patterns across every connected client and server.
2. Why can't I just use my existing APM (like Datadog or New Relic) for MCP server monitoring? Existing APMs often provide MCP support via language-specific SDKs that require code changes and only capture data as generic spans. Spanly is a purpose-built layer that captures the full JSON-RPC payload and works across all languages without modification, giving you deeper, protocol-specific context that your APM lacks for efficient debugging.
3. How quickly can I integrate Spanly with my production MCP server? Integration can be done in under 5 minutes. You can either drop the Spanly SDK into your TypeScript or Python server, wrap any binary with the Spanly CLI command, or deploy it as a Docker sidecar proxy. No schema changes or server rebuilds are necessary to begin monitoring.
4. Does Spanly support data residency requirements for GDPR compliance? Yes. Spanly offers full data residency with dedicated ingest and storage regions in both the US and EU. When you create a project, you select a region, and all telemetry data remains there with no cross-region replication, making it GDPR-friendly by design for EU customers.