Skip to main content

Agents, platforms & MCP

Availability

Open core · self-host + all Splyntra Cloud plans

The Agents section groups three screens that answer "what is running, and how is it behaving?" — registered SDK agents, out-of-process platforms, and MCP servers.

Agents​

The Agents screen monitors your registered agents across environments. Each agent that ingests with a service_name and framework shows up here with its environment, framework label, activity, and risk at a glance. The table is searchable, sortable, and paginated, with Excel export.

Open an agent to reach the agent detail view — its runs, recent traces, error and risk trends, and the environments it appears in. From here you jump straight into a run in Traces.

New agents are added with the Connect an agent wizard.

Agent Platforms​

The Agent Platforms screen lists out-of-process platforms sending telemetry via webhook — Dify, n8n, Flowise, Langflow, Bedrock, Vertex, OpenClaw. Each platform links to a platform / workflow detail view showing its workflows, runs, and status, plus a Connect action that gives you the turnkey webhook URL and an ingest key.

See Agent platforms for how the webhooks work.

MCP Servers​

The MCP Servers screen surfaces every MCP server your agents call, captured from mcp instrumentation. Per server you see latency, failed calls, and flagged (risk-triggering) calls, so a slow or misbehaving tool source stands out across the fleet.

Open a server for the server detail view — its tools, call volume, latency distribution, failures, and flagged calls over time.

See Framework integrations → MCP to instrument MCP.

Next steps​