Pick a benchmark mode and a number of lookups. This works out the tokens those lookups read, with and without jCodeMunch, and what they cost at list price. It's the cost of the lookups themselves, not a forecast of your bill.
Your lookups
A "lookup" = any time the agent reads code to answer something.
Mode assumptions
Modeled cost of the lookups
Tokens per lookup, grep-and-read
—
Tokens per lookup, jCodeMunch
—
Cost per lookup (grep-and-read / jCodeMunch)
—
Cost of — lookups (grep-and-read / jCodeMunch)
—
What this covers: the code the lookups read, at about $5 per million input tokens. A whole session also pays for everything else the agent reads and writes, and that part isn't modeled here.
Honest math:
Both modes are arithmetic on our published benchmark, against an agent that greps and then reads the top files, priced at about $5 per million input tokens (Claude Opus 4.8 list price). "Best Case" is FastAPI, the largest repo we test: 85,296 tokens a lookup for grep-and-read, 2,556 for jCodeMunch. "Benchmark Average" is all 15 tasks on 3 public repos: 44,332 against 1,675. Pick the one that matches your skepticism level.
These are per-lookup figures under benchmark conditions, and the method and raw data are public. They model the lookups, not your bill. Your mileage will vary, which is why every license comes with a 30-day guarantee.
Verified by practitioners
Four signals. Four buyer emotions.
Efficiency
5×
"Roughly 5× more efficient context retrieval."
Artur Skowroński · VirtusLab
Reasoning Quality
Tokens for thinking, not retrieval
"Preserves your context budget for actual reasoning."
Sion Williams
Structural Depth
Queries native tools can't answer
"Structural questions you simply can't ask Grep or Glob."
Traci Lim · Amazon Web Services
Scarcity Economics
Context is the scarce resource
"The whole game is what you choose not to put in the prompt."
Eric Grill
96.2%
fewer tokens per lookup
80%
A/B task success rate (one outside test, one codebase, 25 runs per arm, Sonnet 4.6, March 2026)