keymap — where do your API keys actually live?

In October 2026, a breach at an agent-tool vendor exposed provider API keys (OpenAI, Anthropic, Google) stored in its console workspaces. Developers affected by it faced the same awkward question: which keys do I have, where are they stored, and which ones do I need to rotate?

Most of us can't answer that in under a minute. A provider key gets pasted into a Claude Code config, then a Codex config, then an .env file, then shell history — and every copy is another place it can leak from, another line on the bill if it's abused.

keymap answers the question locally, in seconds:

python -m keymap

It scans the config files your agent tools, CLIs and shells actually read — ~/.claude.json, ~/.codex/config.toml, ~/.env, shell rc files and history, ~/.aws/credentials, and friends — and reports where each provider key sits, whether the same key is sprawled across files, whether its file is world-readable, whether it landed in shell history, and whether it's one careless git add . away from a commit. It finishes with a rotation checklist: the exact revoke pages for every provider it found.

What it never does

• Never prints key material. Reports show only the public prefix (sk-ant-…) plus a SHA-256 fingerprint prefix (…776a5e869d89), so you can recognise the same key across files without seeing it.
• Never sends anything anywhere. No network, no telemetry, no uploads. It only reads files on your machine.
• Never writes. Read-only by design. It tells you what to fix; you fix it.

Demo (all values fake)

examples/demo-home/ is a fake home directory with the same Anthropic key in two files, an OpenAI key in .env and shell history, a GitHub token in ~/.claude.json, and a world-readable .env. Run:

sh examples/run-demo.sh

Real output (see examples/expected-output.txt):

keymap — API key footprint report
Files scanned: 5  Skipped: 0

Keys found: 5 across 5 file(s) — providers: openai, github, anthropic

Occurrences (prefix + fingerprint only — no key material):
  .claude.json:1  github  (unknown)  ghp_…685d57433a9e  [GitHub classic personal access token]
  .codex/config.toml:5  anthropic  api_key  sk-ant-…776a5e869d89  [Anthropic API key]
  .config/app.json:1  anthropic  (unknown)  sk-ant-…776a5e869d89  [Anthropic API key]
  .env:3  openai  OPENAI_API_KEY  sk-proj-…29289fad0238  [OpenAI project API key]
  .zsh_history:1  openai  OPENAI_API_KEY  sk-proj-…29289fad0238  [OpenAI project API key]

Findings:
  [HIGH] world-readable — .env
      File is mode 0o644 — anyone on this machine can read it. Run: chmod 600 .env
  [HIGH] history — .zsh_history
      openai key (…29289fad0238) was typed into shell history. Rotate the key, then clear that history entry.
  [MEDIUM] sprawl — .env, .zsh_history
      Same openai key (…29289fad0238) lives in 2 files. Keep one canonical home for it and delete the copies.
  [MEDIUM] sprawl — .codex/config.toml, .config/app.json
      Same anthropic key (…776a5e869d89) lives in 2 files. Keep one canonical home for it and delete the copies.

Rotation checklist (revoke at the provider, then replace):
  - openai: https://platform.openai.com/api-keys
  - github: https://github.com/settings/tokens
  - anthropic: https://console.anthropic.com/settings/keys

After rotating: delete the old value from every file above, clear it from shell history, and re-run keymap to confirm.

Install

No dependencies — Python 3.10+ standard library only.

git clone https://github.com/techdevfun/keymap
cd keymap
python -m keymap

Or install the keymap command:

pip install .
keymap --help

Usage

python -m keymap                  # scan your home directory
python -m keymap --root ~/work   # scan a different root
python -m keymap .env .codex/config.toml   # scan specific files
python -m keymap --json           # machine-readable output for CI/scripts

Exit codes: 0 = no keys found, 1 = keys/findings found, 2 = usage or I/O error.

What it checks

Check	Severity	Meaning
world-readable	high	File holding a key is readable by anyone on the machine (chmod 600 it)
history	high	Key was typed into shell history — rotate it, then clear the entry
git-tracked	high	Key is already tracked by git — rotate now, history rewrites don't un-leak it
not-ignored	medium	Key sits in a git repo and is not gitignored — one git add . from a commit
sprawl	medium	Same key lives in 2+ files — keep one canonical home, delete the copies

Recognised providers: Anthropic, OpenAI, Google, GitHub (classic, fine-grained, OAuth, server-to-server), AWS access key IDs, Slack, Stripe. Anything assigned to a *_API_KEY / *_TOKEN / *_SECRET / *_PASSWORD variable that matches no known pattern is reported as unrecognized instead of ignored. Placeholder values (your-token-here, ${OPENAI_API_KEY}, xxx…) are skipped.

How it's different

Secret scanners like gitleaks / trufflehog hunt for leaked secrets in repos. keymap answers a different question: the keys you own, on your own machine, right now — which agent tools hold them, how many copies exist, and what to rotate first after a vendor breach. It complements a repo scanner; it doesn't replace one.

Roadmap

• Scan 1Password / macOS Keychain references (names only, no secrets).
• --fix for the safe fixes only (chmod 600, never key deletion).
• Detect keys in agent-tool MCP server configs.
• Spend-anomaly hints: link to each provider's usage page in the report.

License

MIT — see LICENSE.