LanceDB sells an open-core multimodal vector platform; 10 publicly named adopters span media, dev tools, agents and search — a Keenable tie-up is promising but unproven
Asked:
“Please explain what LanceDB sells and who LanceDB's customers are. Is there a mutually beneficial colllaboration between LanceDB and KeenableAI?”
Reviewed on 2026-09-01: 4 official LanceDB product and edition sources, a shortlist of 10 publicly named adopters across 5 customer segments drawn from LanceDB materials, and 2 public descriptions of Keenable's offering. Bar length below counts mentions across reviewed LanceDB sources — a proxy for public prominence, not revenue.
Findings the chart cannot say alone
These are publicly referenced users, adopters and case studies in LanceDB materials — not every named organization is necessarily a current paying Enterprise customer. Only CrewAI, Cognee and CodeRabbit are explicitly labelled case studies.
lancedb.comNetflix reports its Media Data Lake on LanceDB cut p90 latency by over 90%, and CodeRabbit says it kept sub-second responses across millions of code reviews where other vector databases hit cost and performance walls.
lancedb.comEnterprise documentation claims 100B+ row scale, up to 10,000 QPS and 50–200 ms latency from object storage — vendor claims, not independently benchmarked; pricing is not publicly stated in the reviewed pages.
docs.lancedb.comKeenable already integrates with two named LanceDB adopters — CrewAI and AnythingLLM — via its MCP server and agent integrations, a natural entry point for a joint reference architecture.
keenableai.mintlify.app
The open-core product stack
Open source · Apache-licensed
LanceDB OSS
Embedded multimodal / vector database for developers: vector similarity search, full-text search, SQL-style querying over vectors, metadata and raw multimodal data (text, images, videos, point clouds), zero-copy, automatic versioning, GPU-assisted indexing. Runs locally or in your cloud. Stated scale: petabytes of multimodal data, billions of vectors searched in milliseconds; a landing page cites a 20 PB largest table and 20K+ QPS. github.com · lancedb.com
Commercial · managed or BYOC / private cloud
LanceDB Enterprise
Distributed multimodal lakehouse for production retrieval: tiered low-latency cache, horizontal query throughput, distributed search, indexing and compaction, training cache, derived features, backfills, OpenTelemetry metrics and traces, private networking, encryption at rest, SOC 2 Type II and HIPAA coverage. For teams needing more scale, visibility and control than one embedded process. Stated: 100B+ rows, 10,000 QPS, 50–200 ms from object storage. docs.lancedb.com · docs.lancedb.com
A Keenable × LanceDB pilot: live-to-durable research memory
Keenable supplies live web search and content retrieval to agents and AI tools through an MCP server, API and integrations (Cursor, VS Code, Claude Code, CrewAI, AnythingLLM and more). LanceDB supplies the durable multimodal storage, indexing, hybrid vector retrieval and memory layer those agent workflows lack. No public partnership exists today; the fit is architectural, not contractual.
Keenable — live webSearch and fetch the live web via MCP / API; normalize pages, chunks, citations and embeddings.
keenableai.mintlify.app
→
LanceDB — durable memoryStore normalized pages, chunks, citations and embeddings; hybrid vector + full-text + SQL retrieval, tenant/project isolation, provenance, OSS-to-Enterprise deployment path.
→
Agents — query bothAgents query fresh web results and historical project memory together: persistent memory, fewer repeated fetches, enterprise deployment options.
- Value to Keenable: persistent memory, lower repeated-fetch costs, tenant isolation, hybrid retrieval, provenance, enterprise deployment options.
- Value to LanceDB: a high-volume ingestion/retrieval workload, distribution through Keenable's agent integrations, a joint reference architecture and potential Enterprise leads.
- Risks and boundaries: do not duplicate Keenable's web index wholesale; define freshness, deduplication, deletion, copyright/data-retention and tenant-isolation rules; benchmark incremental value against simpler object-storage or relational options.
- Second opportunity: integrate Keenable live search into LanceDB's agent ecosystem and examples, with LanceDB as an optional persistence backend, never a mandatory dependency.
- Verdict: strategically promising, no existing public partnership found — validate with a scoped 4–6 week connector-and-benchmark pilot before any commercial alliance.
Public adopter shortlist
| Organization | Segment | Use case | Relationship wording | Mentions | Source |
|---|
Method: desk review of official LanceDB and Keenable pages on 2026-09-01. 16 rows total: 4 product/edition sources, 10 publicly named adopters (bar length = count of mentions across reviewed LanceDB sources; illustrative shortlist, not exhaustive), 2 Keenable offering descriptions. Scale and latency figures are vendor claims; pricing not publicly stated. Reported-benefit quotes trimmed for space in the table.