// the find
monoscope-tech/monoscope
Monoscope lets you ingest and explore your logs, traces and metrics. We store these in S3 compatible buckets. Query in natural language via LLMs.
Monoscope is an OpenTelemetry-native observability platform (logs, traces, metrics) that stores everything in your own S3 buckets instead of a proprietary data store, backed by their in-house TimeFusion columnar engine. It's built entirely in Haskell, ships a CLI with stable JSON output designed for agentic pipelines, and exposes an MCP server for AI client integration. Best fit for teams who want Datadog-style UX without the Datadog bill, and who can operate a Haskell service.
- S3-native storage is a real architectural win: years of telemetry at object storage prices, no vendor lock-in on the data layer, and you own your own retention policy without paying per-GB ingestion fees.
- The CLI's stable JSON envelope contract (documented output shapes per command, MONOSCOPE_AGENT_MODE flag, cursor-based pagination) shows genuine thought about machine-readable output — most observability CLIs are afterthoughts designed for humans only.
- 99+ numbered SQL migrations checked into the repo is evidence of a real, running production system with real schema evolution, not a polished demo. The migration history is honest about the messy path.
- TimeFusion being PostgreSQL-compatible means standard psql/pgAdmin/DBeaver tooling works against your observability data, which sidesteps the 'proprietary query language' trap that makes migrating away from Elastic painful.
- Haskell is the biggest operational risk for adoption: GHC build times are notorious, the Docker image helps at deploy time, but the first time something goes wrong in production you need someone who can read Haskell stack traces and understand lazy evaluation. Most ops teams can't.
- AGPL-3.0 is papered over in the comparison table as 'Free' — it's not free if you modify and run it as a network service, which is exactly how you'd self-host it. This is a real blocker for many companies and the README doesn't surface it honestly.
- Self-hosted vs cloud feature gap is significant: 'DIY' auth/SSO and 'Basic email' alerts only. If you're self-hosting because you have data residency requirements, you're simultaneously giving up Slack, PagerDuty, and any auth integration — the features most security-sensitive orgs actually need.
- The TimeFusion migration is still in-progress (roadmap item: 'Full migration to TimeFusion storage engine'), meaning the current architecture is in an intermediate state. The repo also has 'timescaledb' as a topic tag, and the migration files contain two different '0016_' files — schema instability signals that the storage layer story isn't settled yet.