Amundsen Data Discovery
Open-source data discovery and metadata engine originally built at Lyft. Amundsen indexes metadata from data sources (tables, dashboards, users) and makes it searchable via Elasticsearch. REST API provides search, table metadata, column lineage, and usage statistics. Uses Neo4j for graph-based lineage. Provides a web UI for data discovery. Apache 2.0 licensed.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
Apache 2.0 open-source — auditable. Default no-auth is a security risk — requires explicit configuration. Self-hosted: security posture depends on deployer. Authentication delegation to OIDC provider when configured.
⚡ Reliability
Best When
You need open-source data discovery for your data lake or warehouse and can operate self-hosted infrastructure (Neo4j + Elasticsearch + Python).
Avoid When
You need managed data catalog with SLA, enterprise support, or integrated data governance — DataHub or Atlan are better maintained and more feature-rich.
Use Cases
- • Search Amundsen's metadata catalog via REST API for datasets matching natural language queries — useful for AI agent data discovery
- • Retrieve table metadata (schema, description, owners, usage frequency) before an agent accesses a dataset
- • Query column-level lineage to understand data provenance before using data for AI model training
- • Build data-aware agent workflows that look up dataset documentation and usage patterns from Amundsen before querying
- • Integrate Amundsen's popularity and usage signals to help agents prioritize high-quality, frequently-used datasets
Not For
- • Enterprises needing enterprise support, SLAs, and managed hosting — Amundsen is self-hosted only
- • Complex data governance workflows — Amundsen is primarily discovery-focused; for policy enforcement use Collibra or BigID
- • Real-time metadata updates — Amundsen relies on periodic ingestion pipelines, not real-time metadata streams
Interface
Authentication
Authentication is optional and configured by the deployer. Supports OIDC/OAuth integration for enterprise deployments. Default setup has no authentication — requires explicit configuration for secure deployments.
Pricing
Amundsen is free and open-source. Operational costs include Neo4j (paid for scale), Elasticsearch, and compute. Self-hosting requires operational expertise.
Agent Metadata
Known Gotchas
- ⚠ Amundsen requires self-hosting Neo4j, Elasticsearch, and multiple Python microservices — significant operational overhead
- ⚠ API documentation is sparse — agents must discover endpoints from source code or community resources
- ⚠ Metadata freshness depends on ingestion pipeline frequency — data may be days old if pipelines run infrequently
- ⚠ No authentication by default — self-hosted deployments without auth are a security risk; always configure OIDC
- ⚠ Amundsen development has slowed compared to DataHub — evaluate long-term maintenance before committing
- ⚠ Search relevance depends on metadata quality — poor table descriptions yield poor search results for agent discovery
- ⚠ Neo4j licensing changed to dual-license (GPL/commercial) — verify licensing implications for your deployment
Alternatives
Full Evaluation Report
Detailed scoring breakdown, competitive positioning, security analysis, and improvement recommendations for Amundsen Data Discovery.
Scores are editorial opinions as of 2026-03-06.