jCodeMunch is a one-time payment, not a subscription. It doesn't change what your AI provider charges. It changes how an agent reads code: one function instead of whole files. On our public benchmark a lookup reads 94.5% fewer tokens on average than grep-and-read (15 lookups on 3 repos, tiktoken cl100k_base). Whether that shows up on your bill depends on the rest of the session, and the 30-day guarantee is there so you can find out.
One developer, commercial use, no renewal.
Up to 5 developers on one payment. When the team grows past five, Platform covers everyone.
Unlimited internal seats. No per-seat invoice surprise next quarter when the team grows.
Paying by invoice? We invoice companies on 30-day terms and take purchase orders: write to j@gravelle.us. For a quote to attach to the request, download the Platform quote (PDF) for jCodeMunch or for all three tools.
Local ONNX embeddings. Indexes never leave the machine. No data egress to a third-party SaaS just to find a function definition.
Server only reads what you whitelist. Path validation, automatic secret redaction, rate limiting.
Apache-2.0 retrieval interface. Vendor-independent. Your investment in retrieval discipline outlives any single tool.
audit_agent_config surfaces exactly what tokens your team is paying for before they reach the model. Hard to budget what you can't measure.
"Roughly 5× more efficient context retrieval."
"Preserves your context budget for actual reasoning."
"Structural questions you simply can't ask Grep or Glob."
"The whole game is what you choose not to put in the prompt."