Wavect · Customer story
Wavect uses coding agents across client and internal projects. The agency measured how LeanCTX changed the context delivered during everyday development.
less context across tracked usage
Tracking snapshot · 26 July 2026Context volume indexed to 100. Observed usage, not a controlled productivity study.
The agency environment
The tracked work spans new builds, inherited systems, bug investigations, integrations and internal tools. Each brings different sources and read depths.
Repeated reads.
Across changing projects.
Each project switch brings new code, schemas and configuration. The same material returns in later reads and command output. Wavect measured how LeanCTX changed the context delivered to its agents.
The operating change
Begin with the smallest useful representation, expand when the decision requires more evidence, and verify against the repository rather than the summary.
Find system shape, likely ownership and evidence needed for the next decision.
Narrow scope with cached context and compact representations.
Inspect exact source before exact edits. Recover raw when meaning is uncertain.
Keep signal from builds, tests and commands. Compress noise, preserve failures.
Review the diff, run the tests, check the real system. Compression does not approve work.
Where the reduction came from
Across files and command output, the overall reduction was 64.1%. MCP tool traffic accounted for most of the saved tokens. Three recurring patterns explain the result.
Compact views helped engineers locate relevant code. Full source returned when the investigation moved to implementation.
Map and signature views reached roughly 97% compression across the tracked usage.
Long sessions revisit the same material. Cached re-reads and delta views reduced how much context had to be delivered again.
This is a share of total savings, not the compression rate of a cache hit.
Builds, tests, Git and package managers produce verbose output. Shell compression kept results and actionable failures while reducing surrounding text.
The highest observed shell result. It is not an average across all commands.
The estimated economics
Across the tracked workload, Wavect estimated a 56.6% reduction in token cost. The input and output figures describe different parts of that estimate.
Pricing, model mix and cache treatment affect what an actual bill looks like. These percentages are not additive.
Workload estimate · not audited invoice ROI
Read the result in context
These results describe tracked context reduction. They do not establish faster delivery, better code or the same savings in every workflow.
The snapshot covers Wavect’s tracked project mix. It is not a controlled A/B experiment, and the report does not establish cycle-time or throughput gains.
The cost figure is a workload estimate. Model prices, cached-input pricing, subscriptions and model mix affect the actual invoice.
Smaller context does not establish correctness or security. Exact edits need original source; acceptance still needs real diffs, tests and review.
Compression varies with file format, repetition and integration. A peak result cannot be applied to every repository or every agency task.
What to take into your evaluation
The strongest fit is repeated access to large repositories or verbose tool output. Measure setup, recovery and review effort alongside token reduction.
Large, repeated workflows — long agent sessions, unfamiliar repositories, recurring reads and verbose tool output.
Small but frequent workflows — individual saving may be modest, but repetition makes the aggregate meaningful.
Large one-off workflows — compression can help, but setup, review and recovery may dominate the economics.
Small one-off tasks — a context layer can add more operational complexity than the task justifies.
The original account
Kevin Riedl’s technical field report describes Wavect’s workflow, recovery rules and evaluation limits in full. The tracking snapshot and the later editorial review are separate dates.
Start with LeanCTX
Start with a familiar repository. Compare the context, recover the detail and review the work.
Local-first. Model-agnostic. Your context.
Cookie settings
Your privacy. Your choice.
We count public page requests without tracking cookies or visitor IDs. Essential only also disables this server statistic for future requests. With your permission, PostHog measures clicks and reading activity. Masked recordings need a separate choice. No advertising.
Essential
Theme, consent choice and account functionality.
Cookieless server statistics
Hourly request counts, without visitor recognition. Your refusal is stored as an essential preference.
Analytics
Optional page analytics · PostHog on our server. Off until you allow it.
Session recordings
Optional replay of public-page interactions, with text and inputs masked. Requires analytics.
Your browser requests privacy protection. Analytics and recordings stay off.
Change your choice anytime in Cookie settings.
