Skip to content

Logchef vs OpenObserve

OpenObserve and Logchef are both AGPL-licensed and both self-hostable, but they are different kinds of software. OpenObserve (O2) is a full observability backend: it ingests logs, metrics and traces and stores them itself as Parquet files. Logchef stores application metadata, but no logs. It is a log UI over the ClickHouse, VictoriaLogs or Grafana Loki that you already run. Choose on that line first, then on the details below.

Pick OpenObserve if you are starting fresh and want one system that takes in logs, metrics, traces and frontend monitoring, keeps them on object storage, and gives you a UI for all of it.

Pick Logchef if your logs already land in ClickHouse, VictoriaLogs or Loki, or you want them there, and you need search, dashboards, alerts, SSO and per-team access control over that store without adding a second one.

Logchef OpenObserve
What it is Query UI and control plane over log storage you already run Observability platform that ingests, stores and queries telemetry (README)
License AGPLv3 Open source edition: AGPLv3. Enterprise edition: commercial EULA (editions)
Storage None of its own. Reads ClickHouse tables, VictoriaLogs and Loki Its own Parquet files on local disk or object storage such as S3, GCS, MinIO or Azure Blob (architecture)
Ingestion None. Your collectors write to the backend Owns it: _json, _multi and _bulk HTTP APIs, OpenTelemetry Collector, Vector, Fluent Bit, Filebeat, syslog and more (ingestion docs)
Query languages LogchefQL, native ClickHouse SQL, native VictoriaLogs LogsQL, native Loki LogQL SQL for logs and traces, SQL or PromQL for metrics, VRL in pipelines
Signals Logs only Logs, metrics, traces, real user monitoring, session replay
Deployment One Go binary, SQLite or Postgres for metadata, plus your backend One binary with SQLite and local disk. HA mode needs Kubernetes, object storage, PostgreSQL and NATS
Auth and RBAC OIDC SSO, local auth, team-scoped source access and an audit log in the open-source release Open source: no SSO, and all users have full access. SSO and RBAC are Enterprise features
Alerting Scheduled alerts with email, webhook, Slack and PagerDuty destinations Scheduled, real-time and composite alerts, email and webhook destinations (alerts docs)
Dashboards Built in: time series, stat, breakdown, table. Scheduled email reports Built in, 19 prebuilt chart types plus custom charts
AI and MCP AI query generation, MCP server with OAuth sign-in AI assistant, MCP server in all editions
Pricing Free. You pay for the infrastructure you run Open source is free. Enterprise is free up to 50 GB of ingestion per day. Cloud lists per-GB ingestion pricing (pricing)

OpenObserve facts are from its docs and repository as of November 2026.

OpenObserve owns the whole path. Collectors send data to its ingestion APIs, ingesters convert it to Parquet, and queriers read those files back. Per its architecture docs, the default is a single node with SQLite and local disk. The HA mode stores all Parquet files in object storage, keeps metadata in PostgreSQL, uses NATS as the cluster coordinator, and does not support local disk storage.

Logchef owns none of that. It connects to a ClickHouse table, a VictoriaLogs instance or a Loki tenant as a source, discovers the fields, and queries it. Beyond a timestamp column it needs no particular schema. Your data stays in a database that other tools can read at the same time: clickhouse-client, Grafana, a BI tool, or a batch job.

The practical result: OpenObserve is less to assemble when you have nothing. Logchef is less to change when you already have ClickHouse, VictoriaLogs or Loki, because nothing gets re-ingested.

OpenObserve covers logs, metrics, traces, real user monitoring and session replay in one UI, per its README. If you want to jump from a slow trace to its logs, it does that.

Logchef does logs. It has no traces UI, no APM, no metrics store and no session replay, and it does not ingest anything. A ClickHouse table can carry trace_id and span_id columns and you can filter on them, but that is filtering, not tracing.

This is the sharpest difference between the two free editions.

OpenObserve’s edition comparison lists single sign-on as not available in the open source edition. Its RBAC docs say: “RBAC is not supported. All users have unrestricted access to all features.” Both are in the Enterprise edition, which is free up to 50 GB of ingestion per day under a commercial license. The comparison page marks SSO and RBAC in Enterprise as requiring HA mode. The RBAC docs also list OpenObserve Cloud.

Logchef’s full access model is in the AGPL release. Users sign in with OIDC or local accounts. Sources are assigned to teams, and a user can query only the sources of their teams. Automation gets scoped service tokens, and teams, sources and memberships can be declared in a TOML file kept in Git. An audit log records configuration changes, sign-ins, queries and data exports for global admins.

Both tools speak SQL, which sets them apart from most log products.

OpenObserve uses SQL for logs and traces, and SQL or PromQL for metrics. Its README makes a point of having no proprietary query language. VRL functions transform data in pipelines.

On a ClickHouse source, Logchef accepts a single SELECT statement, including functions and joins against other tables. It parses the SQL, expands time macros, and applies result limits. VictoriaLogs sources take native LogsQL, and Loki sources take native LogQL. On top of all three sits LogchefQL, a short filter syntax that compiles to each backend’s language. Its aggregations work on ClickHouse and VictoriaLogs.

OpenObserve is built around Parquet on object storage, and its README credits that design for its storage cost. If cheap, long retention on S3 is your main requirement and you have no existing log database, that is a strong reason to choose it.

With Logchef, storage cost is whatever your backend costs. You size and operate that backend yourself. Logchef adds no storage of its own beyond a small metadata database.

  • You want logs, metrics and traces in one product, with correlation between them.
  • You need real user monitoring or session replay.
  • You have no log database yet and want ingestion, storage and UI from one binary.
  • You want object storage as the primary store from day one.
  • You want ingest-time pipelines that parse, enrich or route data with VRL.
  • You want real-time alerts evaluated at ingestion. Logchef alerts are scheduled queries.
  • Your logs are already in ClickHouse, VictoriaLogs or Loki and you do not want to copy them into a second system.
  • You need SSO, per-team access control and an audit log in a fully open-source deployment.
  • You want native ClickHouse SQL over your own tables, including joins with non-log data.
  • You run several backends and want one UI across them, with dashboards that mix ClickHouse, VictoriaLogs and Loki panels.
  • You want a terminal workflow: the CLI runs the same queries with the same permissions, and the command palette runs a search in any source.

There is no dedicated migration guide, and Logchef has no importer for OpenObserve’s Parquet files. The practical path is to re-point ingestion and let old data age out:

  1. Create a log table using the schema design guide, or stand up VictoriaLogs.
  2. Send the same stream there. Vector and the OpenTelemetry Collector are both collectors that OpenObserve also accepts, so a second sink is usually a small config change.
  3. Follow the quickstart, add the source, and run both side by side until you trust the new path.

Metrics, traces and RUM need a home of their own. Logchef will not take them.

For logs, yes, if you run ClickHouse, VictoriaLogs or Loki. For metrics, traces or RUM, no.

No. OpenObserve stores data in its own Parquet layout. Logchef reads ClickHouse, VictoriaLogs and Loki.

Both publish their core under AGPLv3. OpenObserve also ships an Enterprise edition under a commercial license that adds SSO, RBAC, audit trail and federated search. Logchef has one edition, and its SSO, team access and audit log are part of it.

Does OpenObserve open source include SSO or RBAC?

Section titled “Does OpenObserve open source include SSO or RBAC?”

Not per its own docs as of November 2026. They are Enterprise features. Logchef includes OIDC SSO and team-scoped access in its open-source release.

Both, with a difference. OpenObserve’s binary includes storage. Logchef’s binary is the query layer only, and it needs a ClickHouse, VictoriaLogs or Loki backend to read from.

Yes. OpenObserve documents an MCP server in all editions. Logchef’s MCP server is built into the main binary, signs in with OAuth, is read-only, and sees only the teams of the user who approved it.

Ready to try it? Start with the quickstart.