TECHDEV · Open-source tool

keymap

In October 2026, a breach at an agent-tool vendor exposed provider API keys stored in its console workspaces — and every affected developer faced the same awkward question: which keys do I have, where are they stored, and which ones do I need to rotate? 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.

View on GitHub →Read the article →

Python 3.10+ · Standard library only · MIT

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.

Copied verbatim from the keymap README ("What it never does").

The report: a fake-home demo

The repo ships examples/demo-home/, 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. Real output of sh examples/run-demo.sh:

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.

Demo output copied exactly from the keymap README ("Demo (all values fake)" — matches examples/expected-output.txt).

Quick start

# 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

Commands copied exactly from the keymap README ("Install").

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.

Commands and exit codes copied exactly from the keymap README ("Usage").

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

Table copied from the keymap README ("What it checks").

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.

From the keymap README ("Recognised providers").

Honest comparison

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.

From the keymap README ("How it's different").

Roadmap — not shipped yet

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.

Roadmap copied from the keymap README; none of these have shipped yet (2026-10-30).

License

MIT — see LICENSE in the repository.

Have a say

Vote for this tool, suggest a feature, or discuss it with other operators on the community board.