Get 40x more out of every token. Codegraff gives your Claude Code or Codex agent structural file tools — outline-first reads, scope-aware search, 7ms batched ops. Same agent, same workflow, just better I/O underneath.
Less than two days of your Max subscription. Your weekly rate limit goes further.
40x
fewer tokens per read
outline vs full file
6x
faster than ripgrep
hyperfine, 4.7 GB repo
7ms
10 ops in parallel
muonry batch, no fork/exec
none
repeated context
scope dedup across turns
21x
faster than process spawn
0.7ms in-process vs 15ms fork/exec
1.2M
tokens saved
across all users this month
34k
ops this week
batch, search, read, edit
Token usage — last 14 days
hover a bar
tokens consumed per coding session
Search speed — zigrep vs ripgrep
hyperfine · 4.7 GB repo · macOS arm64
A tool upgrade for Claude Code and Codex
One install adds faster read, search, edit, and memory tools via MCP. Your agent uses them automatically. Nothing else changes.
A rate-limit stretcher
Sends outlines, symbols, and ranges instead of entire files — 40x fewer tokens per operation. Your weekly cap goes further.
A persistent daemon, not a bag of scripts
All tools run in-process over a single pipe. 10 parallel ops in 7ms. Cross-session memory deduplicates repeated context automatically.
What this is not
Not a coding agent. Claude does that. Codegraff just makes the file operations underneath faster and leaner.
Not a wrapper or proxy. Your agent talks directly to Anthropic or OpenAI. Codegraff only replaces the file tools.
Not lock-in. MCP-standard tooling. Works with any compatible agent. Uninstall and everything still works — just slower.
What Codegraff adds
Scope mode — context, not noise
Every match returns its enclosing function. Not a line. Not a file. The exact block your agent needs to act on — nothing more.
zigrep -S 'validateInput' src/Cross-search dedup — zero redundant tokens
Scopes already sent to your agent are skipped automatically. Same codebase, 10 turns in — still no repeated context.
zigrep -D 'handleAuth' src/Batch ops — 10 reads in 7ms
Persistent daemon. No fork/exec per call. Reads, searches, diffs, and edits run in parallel over one stdio pipe.
muonry batch [read, search, diff, edit]Structural read — 50 tokens instead of 2000
outline gives you the full symbol map of a file in ~50 tokens. Then symbol pulls exactly the function you need — never the whole file.
zigread outline src/main.zigSymbol-safe patching — no line drift
Edit by function name, not line numbers. Survives upstream diffs, concurrent agents, and mid-session file growth.
zigpatch --symbol parseToken src/auth.zigCross-file atomic replace
Rename a symbol across every file in one pass. Dry-run preview included. No multi-step sed chains.
zigrep --replace 'newName' 'oldName' src/Pricing
Individual
1 seat · all tools · key by email
Team
2 seats · all tools · key by email
Student
90% offEnter your .edu email — we'll send a single-use promo code to apply at checkout.
How it works
curl -fsSL https://codegraff.com/install.sh | shmuonry activate <key> <email>Quick install
# Install all Codegraff tools curl -fsSL https://codegraff.com/install.sh | sh # Activate your license after purchase muonry activate CGPRO-XXXX-XXXX [email protected]
Questions? [email protected] · Terms · Privacy