What is context saving for coding agents?
Context saving is keeping a coding agent's context window for the actual change by serving structural answers — definitions, callers, dependencies, blast radius — from a persistent code graph instead of having the agent re-read source files. Across 955 jobs on 18 public repositories, a typical job took 2,221 tokens in one call, not 13,321 in two.
Where an agent's context goes
A coding agent's context window is its working memory, and most of it is spent on orientation rather than on the change: opening files to find a definition, reading more files to find the callers, reading again to judge what a change would break.
Every re-read costs tokens that are then unavailable for the work, and the structure it rediscovers is thrown away when the session ends.
That cost repeats on every task and every session, because file reading transfers source when the agent only needed an answer about it. A list of callers is a few hundred tokens; the files that contain those callers are tens of thousands.
What a code graph changes
Context saving replaces the re-reading with a query. The repository is indexed once into a code graph — symbols, references, dependencies, effects — and the agent asks it questions over MCP: where is this defined, who calls it, what does it depend on, what breaks if it changes. The answers are computed, so orientation becomes a one-time index cost.
Measured across 18 public repositories of 13,954 files and 2.7 million lines in 11 languages, over 955 repetitions of the job coding agents perform daily — change a function, which needs the function, the code it uses from other files and the code that calls it.
For a typical job, the file-by-file workflow opened 1 file and read 1,413 lines, 13,321 tokens. The same job through Context Zero Engine's code graph used 2,221 tokens in one call: 83% fewer, and 10.1× fewer pooled across the run. Method, numbers and limits are on the research page.
Where the savings come from
The graph answers the questions about structure, and the agent still reads the code it is about to edit. The saving is in the orientation phase, where most of a task's tokens go, and it varies by repository: 2.6× on Express up to 16.5× on Django.