> ## Documentation Index > Fetch the complete documentation index at: https://clickhouse.com/docs/llms.txt > Use this file to discover all available pages before exploring further. # Self-managed ClickHouse Observability > Features, investigation workflows, and operating responsibilities for self-managed ClickHouse Observability > [!NOTE] > > This page covers **self-managed ClickHouse Observability**. Fully managed is recommended for most teams. [Compare the offerings](https://clickhouse.com/docs/clickstack/observability/overview.md#compare-the-offerings) to choose the right option for your team. **Self-managed ClickHouse Observability** is an observability platform for SREs, developers, and their AI agents, built on open-source ClickStack. Search, query, and correlate OpenTelemetry logs, traces, and metrics alongside session replays, with control over ingestion and database configuration. It suits teams that need to customize schemas, tune query performance, or choose how to size and scale their service. Investigate through the ClickStack UI or connect your own agents through the [MCP server](https://clickhouse.com/docs/clickstack/observability/mcp.md). Use search and trace exploration for everyday investigations, and SQL to analyze request details, custom attributes, and business context in your telemetry. ClickHouse Cloud hosts and operates ClickHouse and the ClickStack UI. Self-managed does not require hosting the database or UI yourself. You run the ingestion collector and manage the database configuration for your workload. For most teams, [fully managed ClickHouse Observability](https://clickhouse.com/docs/clickstack/observability/overview.md) is the recommended starting point: send telemetry to a managed OpenTelemetry endpoint, and ClickHouse handles ingestion, schemas, infrastructure, and scaling. Self-managed offers the same ClickStack investigation experience, with responsibility for database tuning and capacity planning. Choose it when you need that control and have the expertise to operate it. ## How data flows Applications and infrastructure send telemetry to your collector. The collector processes it and writes it to your ClickHouse Cloud service. The hosted ClickStack UI queries that data for search, visualization, and investigation. Explore the data in the UI or connect an MCP-compatible agent. Both workflows use the same telemetry. SQL lets you correlate signals, compare historical behavior, and investigate questions beyond predefined views. You remain responsible for instrumenting your applications and collecting telemetry from your environment. See the [self-managed architecture](https://clickhouse.com/docs/clickstack/self-configured/architecture.md) for the component diagram and details. ## What you manage * **Ingestion:** deploy and operate your [OpenTelemetry collector](https://clickhouse.com/docs/clickstack/self-configured/ingesting-data/collector.md), including its receivers, processing, authentication, and connection to ClickHouse Cloud. * **Schemas and data sources:** choose your [tables and schemas](https://clickhouse.com/docs/clickstack/self-configured/ingesting-data/schemas.md) and configure the UI's [data sources](https://clickhouse.com/docs/clickstack/self-configured/managing/config.md) to query them. * **Capacity and performance:** [estimate resources](https://clickhouse.com/docs/clickstack/self-configured/managing/estimating-resources.md), configure sizing and scaling, and [tune performance](https://clickhouse.com/docs/clickstack/self-configured/managing/performance-tuning.md) as your workload changes. * **Retention:** set [data retention policies](https://clickhouse.com/docs/clickstack/self-configured/managing/ttl.md) to balance investigation needs and storage costs. ClickHouse Cloud manages the underlying infrastructure, database upgrades, and backups. This does not remove your responsibility for database tuning and capacity planning. ## Features * Connect your own agents through [MCP](https://clickhouse.com/docs/clickstack/observability/mcp.md) to search telemetry, follow traces, analyze patterns and changes, and run SQL. * Search and correlate logs, traces, metrics, and session replays. Follow requests from browser interactions through backend services and database queries. * Run searches and aggregations across high-cardinality telemetry using [full-text search and Lucene-style filters](https://clickhouse.com/docs/clickstack/observability/features/search.md), such as `level:err`, or SQL. Query JSON fields and custom attributes in logs and trace spans. * Explore telemetry with [AI notebooks](https://clickhouse.com/docs/clickstack/observability/notebooks.md) and create [charts from natural-language questions](https://clickhouse.com/docs/clickstack/observability/text-to-chart.md). * Build [dashboards](https://clickhouse.com/docs/clickstack/observability/features/dashboards/overview.md), configure [alerts](https://clickhouse.com/docs/clickstack/observability/features/alerts.md), and share investigation views through [deep links](https://clickhouse.com/docs/clickstack/observability/url-parameters.md). * Identify recurring failures with [event patterns](https://clickhouse.com/docs/clickstack/observability/features/event-patterns.md), investigate changes with [event deltas](https://clickhouse.com/docs/clickstack/observability/features/event-deltas.md), and live tail logs and traces as new events arrive. * Ingest OpenTelemetry data through a collector you operate, with control over processing, tables, schemas, and retention. ## Principles ### Bring your own agents Connect the agents and development tools your team already uses through MCP. Agents can search logs, follow traces, and identify patterns and changes using the same data available in the UI. They can also run SQL to correlate evidence and explore questions that the investigation tools do not directly cover. ### User-friendly and purpose-built The ClickStack UI brings search, trace exploration, charts, dashboards, and alerts together. Use Lucene-style filters and visual query builders for everyday questions, then turn to SQL when you need more control over the analysis. ### End-to-end observability Connect frontend session replays with application logs, distributed traces, and metrics to follow failures across your system. Wide events keep request details, custom attributes, and business context together in logs and trace spans, helping engineers and agents identify which requests, services, and customers are affected. ### Built for ClickHouse ClickHouse's columnar storage, compression, and parallel query execution support fast searches and aggregations across large volumes of high-cardinality telemetry. Concurrent queries let engineers and agents investigate the same data in parallel. With self-managed deployments, you can tune schemas, ordering keys, and resources for your workload. ### OpenTelemetry-native Send logs, traces, and metrics using standard OpenTelemetry instrumentation, SDKs, and collectors. You operate the ingestion collector and configure how it processes telemetry before writing to ClickHouse. This lets you use existing instrumentation while controlling your ingestion pipeline. ## Get started Follow the [getting started guide](https://clickhouse.com/docs/clickstack/self-configured/getting-started/managed.md) to set up your service and connect a collector. Then use a [sample dataset](https://clickhouse.com/docs/clickstack/observability/example-datasets) to explore the UI; choose the **Self-managed** instructions where available. Before using the service in production, review the [production guidance](https://clickhouse.com/docs/clickstack/self-configured/managing/production.md) and [administration overview](https://clickhouse.com/docs/clickstack/self-configured/managing/overview.md).