Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/andrew-shwetzer/claude-code-saveWrote 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/commands/andrew-shwetzer/claude-code-save/load)<a href="https://agentmods.dev/commands/andrew-shwetzer/claude-code-save/load"><img src="https://agentmods.dev/badge/commands/andrew-shwetzer/claude-code-save/load/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/andrew-shwetzer/claude-code-save/load"><img src="https://agentmods.dev/badge/commands/andrew-shwetzer/claude-code-save/load.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00000 | $0.01455 |
| Opus 5.5 | $0.00000 | $0.00582 |
| Sonnet 5.5 | $0.00000 | $0.00291 |
| Haiku 4.5 | $0.00000 | $0.00145 |
Grade A, and why
load 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 1mo 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 — 148 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Resume from a saved session checkpoint. Reads a structured save state and produces a "Previously on..." briefing so you can pick up exactly where you left off.
Arguments
The user's message after /load determines what to load:
- Nothing (
/load): Load the most recent save for the current CWD's project - Project slug (
/load my-app): Load the most recent save for that project - Save ID (
/load save_1711152000000): Load that specific save
How to execute
Step 0: Sync from remote (if enabled)
Check if ~/.claude/saves/config.json has sync.enabled: true and ~/.claude/saves/ is a git repo. If so, pull latest saves before loading:
cd ~/.claude/saves && git pull --rebase --quiet 2>/dev/null; cd - >/dev/null
This ensures you get saves from other machines. If the pull fails (no network, etc.), continue silently with local saves.
Step 1: Find the save file
If no argument provided:
- Detect project slug from CWD using the same priority as
/save:.claude-save-config.jsonin CWD or parent~/.claude/saves/config.jsonproject mappings- Git repo name
- Directory name
- Find the most recent
.jsonfile in~/.claude/saves/<project-slug>/ - If the directory doesn't exist or is empty:
No saves found for this project. Use /save to create your first checkpoint.
If project slug provided:
- Find the most recent
.jsonfile in~/.claude/saves/<slug>/ - If not found:
No saves found for project "<slug>". Run /saves to see all saved projects.
If save ID provided (starts with save_):
- Search
~/.claude/saves/index.jsonlfor the ID to find the project - Then find the matching JSON file in that project's directory
- If not found:
Save "<id>" not found. Run /saves to browse available saves.
To find the most recent file, list JSON files in the project directory sorted by name descending (filenames are YYYYMMDD_HHMMSS.json, so alphabetical = chronological).
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.
- 1mo ago First seen · 148 lines · 0 tokens per session scan A ca7e6510ab86
load is a command published in the GitHub repository andrew-shwetzer/claude-code-save (2 stars, last pushed 5mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,455 tokens. 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 commands, from other repositories
budget
Manage the token budget for Claude Code sessions. Model-aware — picks the right effective context window and prices per model. The session's REAL model is read from the transcript by the budget hook; config.model is only the fallback. Everything from Sonnet 4.6 up is 1M; Haiku 4.5 is 200K.
orchestrate
Do this task as an orchestrator. The aim is the same result at a fraction of the cost: in a long session, every token in this conversation is re-read on each later request, while a subagent's reading is discarded when it returns. The plugin's subagents run on Haiku 5.5, about 40× cheaper per token than Opus 5.5.
scout-meta-review
Run an interactive Scout meta-review — a system-level audit that sits above the individual session types. Checks whether sessions are running, mistake audit is trending well, proposals are flowing, KB files are healthy, and data-source coverage is consistent across session types. Runs in the current conversation.
scout-work
Interactive work session — walks through today's actionable items one at a time, presents a recommended action with draft content, and executes each one only with explicit approval. Runs in the current conversation (not as a background session).
ccc-status
Turn periodic progress updates for long sessions on or off, choose the channel (Slack, Discord, email or console), interval and detail level, or send one now. Use when the user asks to be kept posted, pinged on progress, or to stop updates.
ccc-teleport
Move work between devices: pull a Claude Code on the web session into this terminal with teleport, or keep a local session controllable from the web or mobile app with Remote Control. Use when switching device, monitoring a long task from a phone, or resuming a cloud session locally.