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.
npx agentmods add commands/joseairosa/recall/setupgit clone --depth 1 https://github.com/joseairosa/recallWhat 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 | $0.00000 | $0.00784 |
| Opus 5 | $0.00000 | $0.00392 |
| Sonnet 5 | $0.00000 | $0.00157 |
| Haiku 4.5 | $0.00000 | $0.00078 |
Grade D, and why
setup scanned grade D with 3 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 2d 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.
Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
chmod 600 ~/.claude/recall/config.json Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
cat > ~/.claude/recall/config.json << EOF Reads MCP configurationmediumAgent snooping
mcp.json carries server URLs and auth tokens; reading it lets a mod discover and abuse other integrations.
- Three separate auth paths exist: hooks read from config.json via lib/config.sh; the plugin .mcp.json reads RECALL_API_KEY from settings.json env; ~/.claude.json is the authoritative runtime override for HTTP MCP server How it starts
The opening of the file, as written. The whole thing — 74 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Configure your Recall API key and server URL. This saves credentials to ~/.claude/recall/config.json (for hooks) and injects RECALL_API_KEY into ~/.claude/settings.json (for the plugin MCP server) so both systems authenticate correctly.
Instructions
- Ask the user for their Recall API key. It starts with
sk-and can be found at https://recallmcp.com/dashboard/keys - Ask if they want to use the default server (https://recallmcp.com) or a custom self-hosted URL
- Write the config file (used by lifecycle hooks via lib/config.sh):
mkdir -p ~/.claude/recall
cat > ~/.claude/recall/config.json << EOF
{
"api_key": "THE_API_KEY",
"server_url": "THE_SERVER_URL"
}
EOF
chmod 600 ~/.claude/recall/config.json
- Inject RECALL_API_KEY into ~/.claude/settings.json env section (belt-and-suspenders for hooks that read from env):
python3 - <<'PYEOF'
import json, os
settings_path = os.path.expanduser("~/.claude/settings.json")
with open(settings_path) as f:
settings = json.load(f)
settings.setdefault("env", {})["RECALL_API_KEY"] = "THE_API_KEY"
with open(settings_path, "w") as f:
json.dump(settings, f, indent=2)
print("RECALL_API_KEY written to settings.json env section")
PYEOF
- Update ~/.claude.json so the authoritative runtime config also has the new key. Claude Code reads ~/.claude.json as the final source for HTTP MCP server headers — if an entry for recall-remote already exists there, it bypasses the settings.json env var entirely:
python3 - <<'PYEOF'
import json, os
path = os.path.expanduser("~/.claude.json")
try:
with open(path) as f:
config = json.load(f)
except FileNotFoundError:
config = {}
mcp = config.setdefault("mcpServers", {})
if "recall-remote" not in mcp:
mcp["recall-remote"] = {"type": "http", "url": "THE_SERVER_URL/mcp", "headers": {}}
mcp["recall-remote"].setdefault("headers", {})["Authorization"] = "Bearer THE_API_KEY"
with open(path, "w") as f:
json.dump(config, f, indent=2)
print("recall-remote updated in ~/.claude.json")
PYEOF
- Tell the user to restart Claude Code once — the MCP server reads env vars at launch, so the new key takes effect on next start.
- After restart, verify the connection by calling
mcp__recall-remote__get_workspace. If it succeeds, Recall is working. - Set up the workspace for the current project by calling
set_workspacewith the current directory and git remote.
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.
- 2d ago First seen · 74 lines · 0 tokens per session scan D 311183ee23d7
setup is a command published in the GitHub repository joseairosa/recall (175 stars, last pushed 2d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 784 tokens. A static security scan graded it D with 3 findings (asks for root, reads agent configuration directories, reads mcp configuration). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
debug
Systematic debugging with hypotheses and evidence gathering.
context-stats
Show context window usage, active tier, and system stats.
think
Apply structured thinking frameworks to a problem.
whats-next
Show current project status and suggest next steps.
recover
Scan orphaned worktrees and stale branches, then offer explicit approval-gated cleanup.
capture
Triage raw inbox notes into reviewed repository destinations without deleting their sources.