StackMap
Subscribe

ripwire vs tokensave

Red Hat's 'ripgrep of AI context': a zero-dependency C++23 CLI + MCP that hands a coding agent a ranked call graph — what to touch, blast radius, tests to run — with no index server. — versus — Code-intelligence MCP server for coding agents — a pre-indexed semantic graph (libSQL + FTS5) they query instead of grepping: symbols, callers, impact radius in one call. 100% local, 50+ languages.

The curated verdict

Same goal — replace grep-and-read-whole-files with a pre-indexed graph query — but tokensave keeps a libSQL+FTS5 index behind an MCP server while ripwire recomputes per run with no store.

ripwiretokensave
Stars1.9k624
Forks11163
LanguageC++Rust
LicenseApache-2.0MIT
Last activityyesterday3 days ago
Topicscode-intel, codingcode-intel, local
Curated connections716

ripwire — the curator's take

Reach for this when you want code-context for an agent with the operational profile of ripgrep: one static binary, indexes a repo in well under a second on a few MB, no daemon, no embeddings, no API key — and it ships skills that tell Claude Code/Codex/Cursor/aider *when* to call it. The `--for="<change>"` mode (rank by intended change, then blast radius + tests-to-run + McCabe/Halstead quality deltas) is the differentiator over plain symbol graphs. Prefer the CLI; the MCP schema costs context every session. Not for you if you need a persistent, queryable graph across repos, natural-language-to-Cypher, or semantic search — it is deterministic and lexical/structural only.

tokensave — the curator's take

Use it to stop agents burning tokens on grep/glob/read exploration: one MCP call returns the symbols, relationships and snippets a task needs, and it's local (libSQL, no cloud). Broad reach — 50+ languages, 12+ agent integrations. NOT worth it on small repos where a couple of greps suffice, and it's an index you must keep fresh (re-index on change) — a stale graph misleads the agent. 80+ tools is a lot of surface; most tasks touch a handful.