Splyntra
Get demo
Observability · Tools & RAG

Deep visibility into agent tools, MCP & RAG retrieval

Autonomous agents live or die by their tools and retrieved context. Splyntra validates tool call arguments, inspects return payloads, benchmarks Model Context Protocol (MCP) servers, and measures RAG vector chunk relevance.

Start free on this feature →View documentationBook technical demo
tool_span.json
json
{
  "span_name": "tools.execute.sql_query",
  "tool_type": "mcp_server",
  "server_name": "postgres-mcp",
  "arguments": { "query": "SELECT balance FROM accounts WHERE id = ?" },
  "duration_ms": 142,
  "status": "OK",
  "risk_flags": 0,
  "bytes_returned": 2048,
  "context_relevance_score": 0.94
}
JSON Schema
Payload Validation
Detect malformed tool inputs/outputs.
MCP Native
Protocol Support
Inspect Model Context Protocol calls.
RAG Relevance
Embedding Quality
Evaluate retrieved chunk precision.
< 1ms
Payload Profiler
Zero lag argument & return inspector.

Engineered for high-throughput autonomous agents

Every capability is built into the OpenTelemetry streaming pipeline with sub-millisecond ingestion overhead.

01Tool Profiling

Tool Execution & Parameter Inspection

Inspect the exact JSON parameters passed by the LLM to functions, external APIs, and database tools, along with response payloads.

  • ✓Detect schema violations and malformed arguments automatically
  • ✓Track tool call latency vs execution success rates
  • ✓DLP check on tool arguments to prevent secret leakage to third-party APIs
02MCP Servers

Model Context Protocol (MCP) Observability

Native monitoring for MCP client/server architectures. Track tool schemas, transport latency, and security permissions.

  • ✓Monitor MCP server health, connection stability, and tool discovery
  • ✓Intercept unauthorized MCP server calls before execution
  • ✓Full trace telemetry for local and remote MCP integrations
03RAG Intelligence

RAG Retrieval & Vector Quality Metrics

Evaluate vector search quality, retrieved context window saturation, embedding latency, and chunk hallucination rates.

  • ✓Chunk relevance scoring and context utilization percentage
  • ✓Pinecone, Qdrant, Weaviate, Milvus, and pgvector integration
  • ✓Identify poison documents and noisy context degradation

Frequently Asked Questions

How does Splyntra intercept tool calls?▼
The Splyntra SDK instruments standard tool execution pipelines in LangGraph, CrewAI, OpenAI Agents, LlamaIndex, and MCP clients, recording inputs, outputs, errors, and timing metadata as OpenTelemetry spans.
Does Splyntra work with local MCP servers?▼
Yes. Splyntra supports both stdio and SSE transports for Model Context Protocol servers.

Related Platform Features & Guides

ObservabilityTraces & Replay →IntegrationsMCP Integrations →SecurityRisk Score & Prompt Defense →

Ready to monitor and secure your AI agents?

Get started in under 3 minutes with zero credit card required. Free tier includes up to 5 projects and community telemetry.

Start Free on CloudRequest Custom Enterprise Demo