Gradiva.

Technical notes

Utopik Core reference-host memory measurement.

Conditions and results for the synthetic memory workload cited on the Utopik Core page.

01

Metric

Combined peak RSS is the reference-host process plus all live run-command child processes sampled in the same measurement cycle; host RSS is read immediately before the descendant-process snapshot. Each concurrency point ran in a fresh process and was repeated five times sequentially on 3 September 2026.

Host + live children
02

Workload

Each simulated-model session used a 150 ms response delay and made real read, write and node-version command calls. The workload did not include real model-provider traffic, MCP servers, streaming or multi-turn conversations.

Synthetic tool workload
03

Results

SessionsMean observed peak RSSSDObserved range
10135.64 MiB18.25 MiB120.25–157.39 MiB
30138.30 MiB7.66 MiB129.80–146.61 MiB

Five sequential fresh-process repeats per point. The 30-session point was sampled every 23.2–24.9 ms.

Reference host + live children
04

Interpretation

These are reference-host memory results for this workload. Hardware, operating-system, architecture and exact Node-version details were not recorded in the benchmark receipt, so this is a dated memory observation rather than a reproducible cross-machine performance comparison. Short-lived command processes can start and exit between samples, so sampled child counts and combined peaks can miss brief peaks.

Measurement boundary
05

Dependency boundary

@utopik/core declares one direct runtime dependency: the official MCP SDK. That SDK brings its own transitive dependencies. Core uses no model-provider SDK.

Runtime integration
© 2026 Gradiva Limited.