https://mcpd.sh/blog Articles on tracing, testing and running AI agents with mcpd. en Fri, 06 Nov 2026 00:00:00 GMT https://mcpd.sh/blog/agent-observability-platformhttps://mcpd.sh/blog/agent-observability-platformFri, 06 Nov 2026 00:00:00 GMTWhat is an open source agent observability platform? It records each run as a tree of steps with prompt, completion and cost. Start with stock OpenTelemetry.agent observabilityopentelemetryopen sourcellm cost trackingself-hostingagent evaluation https://mcpd.sh/blog/llm-cost-per-projecthttps://mcpd.sh/blog/llm-cost-per-projectThu, 05 Nov 2026 00:00:00 GMTHow do you calculate LLM cost per project? Price each call from its usage fields, sum the runs, and compare with traces. Learn to cap spend with a monitor.llm costtoken pricingprompt cachingagent observabilitycost monitoringopentelemetry https://mcpd.sh/blog/self-hosted-llm-tracing-postgreshttps://mcpd.sh/blog/self-hosted-llm-tracing-postgresWed, 04 Nov 2026 00:00:00 GMTHow do you run self hosted LLM tracing on Postgres? Run one binary, set its storage to Postgres and send OpenTelemetry spans. Learn the setup and start free.llm tracingpostgresopentelemetryself hostingobservabilitydata retention https://mcpd.sh/blog/single-binary-agent-monitoringhttps://mcpd.sh/blog/single-binary-agent-monitoringTue, 03 Nov 2026 00:00:00 GMTWhat is single binary agent monitoring? One executable with built-in storage and UI that traces every model and tool call. Learn setup, monitors and limits.agent monitoringopentelemetrysqlitellm observabilitycost monitoring https://mcpd.sh/blog/llm-audit-loghttps://mcpd.sh/blog/llm-audit-logMon, 02 Nov 2026 00:00:00 GMTWhat goes in an LLM audit log? One trace per run holds every model call, tool call and its cost. Learn to record, redact and retain it with OpenTelemetry.llm audit logopentelemetryai agentsdata retentionredactionaccess control https://mcpd.sh/blog/agent-workflow-state-trackinghttps://mcpd.sh/blog/agent-workflow-state-trackingSun, 01 Nov 2026 00:00:00 GMTWhat is agent workflow state tracking? Record the cursor, retries, waits and payloads as OpenTelemetry traces. Learn to trace a stuck run in any framework.agent observabilityopentelemetrydurable workflowsai agentstracing https://mcpd.sh/blog/llm-evaluation-datasethttps://mcpd.sh/blog/llm-evaluation-datasetSat, 31 Oct 2026 00:00:00 GMTWhat goes into an LLM evaluation dataset? Frozen cases from your own production runs, each with input, expected behaviour and metadata. Learn to build one.llm evaluationevaluation datasetai agentsllm-as-judgeregression testingobservability https://mcpd.sh/blog/start-tracing-ai-agents-freehttps://mcpd.sh/blog/start-tracing-ai-agents-freeFri, 30 Oct 2026 00:00:00 GMTHow do you start tracing AI agents free? Set two OpenTelemetry variables and a project key, run your agent, and see each prompt, answer and cost. Start now.ai agentstracingopentelemetryobservabilityllm cost