Skip to main content

Maximal token-efficient RAG for headless Claude. Uses your existing claude CLI; auth-agnostic; slice-level retrieval.

Project description

jragmunch-cli

Maximal token-efficient RAG for headless Claude. Uses your existing claude CLI; auth-agnostic; slice-level retrieval powered by jcodemunch-mcp.

Billing: subscription by default, API on opt-in

By default, jragmunch never bills your Anthropic API account. It strips ANTHROPIC_API_KEY and ANTHROPIC_AUTH_TOKEN from the subprocess environment before spawning claude, so the CLI uses your Max / Pro Claude OAuth login while respecting their TOS — you pay $0 in dollars, the work counts against your subscription's session limits.

If you want to bill via the API instead, pass --use-api:

jragmunch --use-api ask "..."

Every verb prints the cost split:

[tokens in=24 out=1273  cost actual=$0.0000 (notional=$0.5334, auth=subscription)  time=27549ms]
  • actual — what you were really billed (always $0 in subscription mode).
  • notional — what the work would have cost via the API. claude -p computes this regardless of auth mode; we surface it as a "what it might have cost" yardstick.
  • authsubscription or api. Run jragmunch doctor to see your resolved mode.

Why

Headless Claude (claude -p) is the right substrate for code automation — CI bots, batch refactors, fan-out agents, internal "chat with your repo" services. The default pattern is "stuff the relevant files into the prompt and pray," which burns tokens on code the model never needed.

jragmunch wraps claude -p with jcodemunch pre-wired so the model retrieves slices on demand instead of receiving giant context dumps.

Install

pip install jragmunch
jragmunch doctor

Requires the claude CLI on PATH (npm install -g @anthropic-ai/claude-code) and jcodemunch-mcp registered as an MCP server.

Usage

jragmunch ask "how does auth work in this repo"
jragmunch ask "what does AuthMiddleware.verify do" --json
jragmunch index --repo .
jragmunch run "Refactor the rate-limiter to use a token bucket"

Verbs (v0.1)

Verb Status Purpose
doctor shipped Verify claude + MCP wiring
ask shipped Retrieval-augmented Q&A
index shipped Index a repo via jcodemunch
run shipped Power-user prompt passthrough
review shipped Diff-aware PR review
changelog shipped Summarize changes since tag
refactor shipped Fan-out batch refactor
tests shipped Generate tests for untested symbols
sweep shipped Pattern-driven cleanup

See PRD.md for the full product spec.

Principles

  • Auth-agnostic. Whatever auth the local claude binary uses, jragmunch uses.
  • Slice, don't dump. Default behavior is jcodemunch retrieval.
  • Structured output. Every verb returns JSON with citations and _meta (tokens, cost, wall time).
  • Composable. --print-command shows the exact claude -p invocation that would run.

License

Apache 2.0

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

jragmunch-0.4.3.tar.gz (22.7 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

jragmunch-0.4.3-py3-none-any.whl (26.0 kB view details)

Uploaded Python 3

File details

Details for the file jragmunch-0.4.3.tar.gz.

File metadata

  • Download URL: jragmunch-0.4.3.tar.gz
  • Upload date:
  • Size: 22.7 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.14.2

File hashes

Hashes for jragmunch-0.4.3.tar.gz
Algorithm Hash digest
SHA256 00296bd2e81129549002c1f3a495415818b44489986627cd596d97ce398c8640
MD5 415133597e8867830f69bb3bc73a2a4e
BLAKE2b-256 32ba40363acb8f1b7b4302f6724d85489c4e05b4d49146e220cf071267dd2347

See more details on using hashes here.

File details

Details for the file jragmunch-0.4.3-py3-none-any.whl.

File metadata

  • Download URL: jragmunch-0.4.3-py3-none-any.whl
  • Upload date:
  • Size: 26.0 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.14.2

File hashes

Hashes for jragmunch-0.4.3-py3-none-any.whl
Algorithm Hash digest
SHA256 8fad93f72412dbae175db743a7f29bc90dc070303f9cffc5383a2b5ac42a291c
MD5 1c184e2435fffeb045e415913ea37c05
BLAKE2b-256 19539570b76e3805357eb5dac26eca1c95ec5ce9ae0b892c78e1c9a2254177a3

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page