// A reference for Claude Code's agent harness architecture, based on a clean-room Rust reimplementation. Built for infrastructure operators who want to understand how Claude Code works under the hood, optimize token usage, and run offline wor
claude-code-internals
git log --oneline --stat
HEAD
- Stars
- 1
- Forks
- 3
- Updated
- Apr 21, 2026
repo --stat
stars
1
forks
3
last update
Apr 21, 2026
license
MITv1.0.0
quickstart.sh
3 steps
- Install
// Drops SKILL.md into ~/.claude/skills/
$ claude skills add stevenjvik-claude-code-internals - Invoke
// Run from any project directory
$ claude --skill stevenjvik-claude-code-internals "fine-tune on this CSV" - Or copy this from the README
// extracted from a Usage / Example section
$ rust/crates/ api/ Anthropic HTTP client, SSE streaming, OAuth commands/ Slash command registry (/status, /compact, /model, etc.) compat-harness/ TS manifest extraction (for parity auditing) runtime/ The brain: conversation loop, config, sessions, MCP, prompts, hooks rusty-claude-cli/ The face: REPL, CLI, streaming display, tool call rendering
stevenjvik-claude-code-internals/
references
- references/
- SKILL.mdopen
- README.mdopen
SKILL.md
readonly
- name:
- claude-code-internals
- slug:
- stevenjvik-claude-code-internals
- version:
- v1.0.0
- license:
- MIT
- author:
- @stevenjvik
- repository:
- github.com/stevenjvik/claude-code-internals
- categories:
- tags:
- #agent-harness#agents#anthropic-claude#claude-code#llm-tools#mcp#rust#token-optimization
- description:
A reference for Claude Code's agent harness architecture, based on a clean-room Rust reimplementation. Built for infrastructure operators who want to understand how Claude Code works under the hood, optimize token usage, and run offline wor
README.md
stevenjvik-claude-code-internals/README.md
12 sections
Loading README…
$ cat reviews/
Reviews
// No reviews yet. Be the first.
Loading review form…