ObsessionDB
Managed ClickHouse for AI agents, with isolated compute so they can't slow prod.
ObsessionDB is managed ClickHouse with compute and storage split apart, so an AI agent probing your data runs on its own isolated, read-only compute instead of competing with production traffic. A cache mesh keeps hot data off the S3 round-trip, and you're billed for compute hours rather than per query, so an agent asking "100x more questions" doesn't inflate the bill. It connects over MCP to Claude, Cursor, Codex, and other agent tools, and claims sub-400ms p99 at 18TB with 20-40% lower cost than other managed ClickHouse providers.
What holds up
- +Compute-compute separation gives each agent its own isolated compute, so it can't slow production queries
- +Billed by compute hours, not per query, so an agent's hundreds of probes don't add a per-query tax
- +Every agent query is logged in full and replayable, no black box reaching into production
Mind the limits
- −Agent sandboxing per branch is still early access, not generally available yet
- −Built specifically around ClickHouse and agent workloads, not a fit if you need a different database engine
Picked here? Add the badge
Put it on your site. It links back to this page. Free for every listed product.
<a href="https://stillworks.dev/products/p/obsessiondb/"><img src="https://stillworks.dev/badge/obsessiondb.svg" alt="Picked by StillWorks" width="230" height="54"></a>