Borrowing it
Nothing to install: this file belongs to sakebomb/mcp-recall. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/sakebomb/mcp-recall/main/CLAUDE.mdgit clone --depth 1 https://github.com/sakebomb/mcp-recallWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/sakebomb/mcp-recall/claude-md)<a href="https://agentmods.dev/instructions/sakebomb/mcp-recall/claude-md"><img src="https://agentmods.dev/badge/instructions/sakebomb/mcp-recall/claude-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.02988 | $0.02988 |
| Opus 5 | $0.01494 | $0.01494 |
| Sonnet 5 | $0.00598 | $0.00598 |
| Haiku 4.5 | $0.00299 | $0.00299 |
Grade A, and why
mcp-recall CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 182 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md – mcp-recall
Project-level conventions. Global ~/.claude/CLAUDE.md guardrails always take precedence on conflicts.
Project Overview
mcp-recall is a Claude Code plugin that intercepts large MCP tool outputs (Playwright snapshots, GitHub API responses, large file reads), compresses them, stores full versions in SQLite with FTS, and delivers brief summaries to Claude. When Claude needs detail, it retrieves via recall__* tools. Goal: enable 3+ hour sessions by preventing context window exhaustion.
Tech Stack
- Runtime: Bun (package manager + test runner — use
bunnotnpm) - Language: TypeScript strict mode, ESNext target
- Database: SQLite with FTS5 (built into Bun via
bun:sqlite) - Schema validation: Zod
- Config format: TOML (
smol-toml)
Commands
bun test # run all tests
bun test --watch # watch mode
bun run typecheck # tsc --noEmit
bun run build # bundle src/ → plugins/mcp-recall/dist/
bun run start # MCP server
bun run dev # MCP server in watch mode
No just / make in this project. Use bun test directly (not just test).
Architecture
bin/recall Entrypoint used by hooks/hooks.json (session-start, post-tool-use)
and as the npm `bin`. Resolves symlinks before locating src/.
src/
server.ts MCP server — exposes the recall__* tools listed below
cli.ts Hook CLI dispatcher (mcp-recall install / learn / profiles / …)
config.ts TOML config loader (Zod-validated, cached)
project-key.ts Git root detection + SHA256 path hash
log.ts Unified stderr logging — info/warn/error/debug, gated on RECALL_DEBUG=1
format.ts Shared byte-size and relative-time formatting utilities
hints.ts Extracts salient search terms for the summary header (deterministic)
tools.ts Tool handler logic for all recall__* tools (pure functions; server.ts wires to MCP SDK)
denylist.ts Built-in + configurable denylist
secrets.ts Secret pattern detection before any write
db/
types.ts All TypeScript interfaces for the DB layer
schema.ts DDL, migrations, getDb / closeDb / initSchema / defaultDbPath
chunking.ts CHUNK_SIZE, chunkText(), sanitizeFtsQuery()
queries.ts Core CRUD — storeOutput, retrieveOutput, evictIfNeeded, forgetOutputs, …
analytics.ts Aggregation queries — getStats, getContext, getSessionSummary, getSuggestions, …
index.ts Re-export barrel (all db/* in one import surface)
handlers/ Compression handlers. The 7-step dispatch order is documented
canonically above getHandler() in index.ts — read it there rather
than trusting a paraphrase. The consequential part: user and
community profiles BEAT the TypeScript registry, bundled profiles
LOSE to it, and native Bash is routed first, before any of them.
Counting handlers is ambiguous (registry entries vs files vs
fallbacks), so HANDLER_REGISTRY is the source of truth.
hooks/ Hook implementations (SessionStart, PostToolUse)
gc/ mcp-recall gc — classifies every project DB and reclaims disk.
Deletion policy lives in STATUS_POLICY; see the safety note below.
import/ mcp-recall import — restores a recall__export dump into a store
install/ mcp-recall install / uninstall / status — writes to ~/.claude.json, settings.json, CLAUDE.md
learn/ mcp-recall learn — reads ~/.claude.json, spawns MCP servers via stdio, generates TOML profiles
profiles/ mcp-recall profiles subcommands (list, install, update, remove, seed, feed, check, test, info, available, retrain)
tests/ Bun tests, co-located by module name
.claude-plugin/ Root plugin manifest (local dev / manual install)
hooks/hooks.json Hook definitions — canonical source, copied to plugins/ on build
plugins/mcp-recall/ Marketplace-installable plugin bundle
dist/ Bundled server.js + cli.js (bun build --target bun)
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago Changed · +65 tokens per session 3b5e2db5be32
- 8d ago First seen · 182 lines · 2,923 tokens per session scan A 354ab614f61a
mcp-recall CLAUDE.md is an instructions file published in the GitHub repository sakebomb/mcp-recall (8 stars, last pushed 3d ago), licensed MIT. It adds 2,988 tokens to every session, about $0.0149 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.