Klavex v0.1.4
Secrets manager · CLI-first · solo & teams

Keep your secrets off disk — and out of your agents' reach.

Klavex injects your environment variables into your process at runtime, so there's no .env for Cursor or Claude Code to read. Import your existing .env in one command — no rewriting. Free for solo devs, one flat price for teams, and every AI agent gets a read-only token scoped to only the envs you pick.

Free for solo devs — no card · pip install klavex · macOS, Linux, WSL
~/repos/checkout-api · zsh
klavex init
found .env — imported 7 secrets into checkout-api/Production
.klavex written · tip: delete the local .env
klavex run -- node server.js
injecting 7 secrets into child process
Server listening on :3000
# meanwhile, your agent goes looking for secrets…
cat .env
cat: .env: No such file or directory
# nothing on disk. the keys only ever lived inside that one child process.
Built for teams shipping with
Cursor·Claude Code·Copilot·Aider·Continue·MCP servers
The agent problem

Your .env is a coding-agent buffet.

It might be in .gitignore, but it's still sitting plaintext on your disk — one read_file('.env') tool call away from a model's context window. And sharing it with the team is still a Slack DM.

Today

Agents read everything. Teams sync nothing.

The same .env that lets your dev server boot also feeds every coding agent in your editor — and getting a teammate set up is still a manual copy-paste ritual.

×Agents have full read access. Cursor, Claude Code, and MCP servers can cat your .env any time — there's no way to scope them down.
×Onboarding takes a day. New hire? Ping six people on Slack for the right keys, paste them into a fresh .env, hope nothing's stale.
×Drift between machines. Someone rotates a key. Half the team's local .env is now broken — and nobody knows why CI keeps passing.
×No audit, no revocation. When someone leaves, every key they ever touched is theoretically still on their laptop.
With Klavex

One source of truth. Nothing on disk to leak.

One encrypted vault — for you, or shared with the team. The CLI pulls the right secrets straight into the process you're running — they're never written to a file an agent can open.

Bring your existing .env. klavex init finds your .env and imports every var in one go — no retyping. Then delete the file and you're secured.
Scope agents, don't trust them. There's no .env to grab — and when an agent does need keys, you mint it a read-only token scoped to specific environments. Dev yes, prod never.
Team sync in seconds. Invite by email — they pip install klavex, run klavex login, and every repo is already there.
Rotate once, everywhere. Update a key in the dashboard. Every shell on every machine picks it up on next klavex run.
Audit + instant revoke. Every fetch is logged. Someone leaves? One click, every machine running their token loses access.
How it works

Set up in minutes. Zero .env files.

Install, log in, point it at your repo once. Then every command you'd normally run gets klavex run -- in front of it.

STEP 01

Install & log in

One Python package, then a browser-based device login that binds the CLI to your machine. No long-lived API keys to leak.

$ pip install klavex $ klavex login # ✓ authorized as sam@team
STEP 02

Init & import your .env

Run it in your repo. klavex init finds your existing .env and imports every var — no retyping. Then delete the file.

$ klavex init # → imported 7 secrets ✓ # → tip: delete the local .env
STEP 03

Run anything

Wrap your dev server, test runner, deploy script — anything. Secrets are injected only into that one child process, never to disk.

$ klavex run -- npm start # ✓ 7 secrets injected
For AI coding agents

Give your agents the keys you choose — and nothing else.

Your secrets aren't sitting in a .env for an agent to scrape. And when you do want Cursor, Claude Code, or a CI runner to have some, mint it its own token — read-only, scoped to the exact environments you pick. It reads Dev; Production stays invisible.

  • Scoped to the environments you grant. The API refuses to decrypt anything outside an agent's allowlist — grant Dev, and Production simply isn't reachable.
  • Read-only by design. An agent token can read its scoped secrets but can't create, change, or delete a thing.
  • No .env, no keychain. The agent authenticates with KLAVEX_TOKEN and runs klavex run — nothing written to disk.
  • Logged and revocable. Every fetch lands in the audit log; revoke the token in one click and it stops working everywhere.
  • Unlimited, and free. Agents never count against your seats — mint one per repo, per CI job, per teammate's editor. You're only billed for human teammates.
checkout-api · team
SC
Sam Chen (you) owner · all environments
OWNER
Cursor agent · read-only
Dev
CI · GitHub Actions agent · read-only
Staging
Cursor read checkout-api/Dev · 46s ago. Production? Never granted.
What's inside

Everything you need. Nothing you don't.

The essentials done well — strong encryption, environments, audit, rotation, scoped access. No sprawling platform to operate, no SDK to wire into your app, no feature bloat to learn.

AES-256-GCM encryption at rest

Envelope-encrypted with a KMS-protected master key. Encryption context binds every ciphertext to its team and repo, so a stolen token can't unlock anyone else's data.

Environments as first class

Production, Staging, Dev — and any custom env you need. Same keys, different values, the right one injected per run.

Audit log for everything

Every fetch and every membership change — by actor, IP, and timestamp. 7-day retention on Solo, 90-day on paid plans.

One-shot rotation

Rotate a key in the dashboard. Every shell, every CI runner, every agent picks it up on next exec. No redeploys.

Team scoping that fits

Owner / Admin / Editor / Viewer roles across your team. Invite teammates by email, change a role or revoke access in a click.

Drop-in for CI & Docker

klavex run works in GitHub Actions, GitLab CI, and Docker. Locally it's klavex run -- <cmd>; in CI a scoped KLAVEX_TOKEN injects the variables into that one process only.

Why Klavex

The simple one. Built for solo devs and small teams.

Doppler and Infisical are powerful platforms — and overkill if you just want your secrets out of .env. Klavex does the one thing well: three commands, nothing on disk, one flat price.

Klavex Doppler Infisical
Pricing model Flat per team — agents free Per seat (~$21/user) Per identity (~$18)
Cost at 10 people $29/mo flat ~$210/mo ~$180/mo
Setup 3 commands, no platform Hosted, polished UX A platform to run / self-host
Local .env on disk Never — no file mode Optional (can inject) Optional (can inject)
Import your .env One command, no rewriting Yes Yes
Scoped AI-agent tokens Unlimited, free, per-env Service tokens (per seat) Yes — incl. Agent Vault

Need self-hosting, a full platform, or proxy-grade agent isolation? Infisical is the stronger pick — and we'll say so. Klavex's bet is narrower: the simplest setup, nothing on disk, one flat price. Read the full comparison →

Pricing

Free to start. Flat team pricing.

Every plan includes unlimited repos, environments, variables — and unlimited AI agents, free. You only pay for human teammates, and the price is flat — not per seat. Save ~17% annually.

Solo

$0/forever

For one developer. Everything you need to never write a .env again — no card, no trial clock.

  • 1 developer (you)
  • Unlimited AI agents, free
  • Unlimited repos, envs, vars
  • Import your existing .env
  • 7-day audit retention
Most popular

Team

$29/mo

For small teams sharing secrets across repos. $24/mo billed annually — save ~17%.

  • Up to 10 teammates — flat price
  • Unlimited AI agents, free
  • Unlimited repos, envs, vars
  • 90-day audit retention
  • AES-256-GCM with KMS
  • Priority email support

Enterprise

$59/mo

For larger teams running Klavex across multiple squads. $48/mo billed annually — save ~17%.

  • Up to 100 teammates — flat price
  • Unlimited AI agents, free
  • Unlimited repos, envs, vars
  • 90-day audit retention
  • Bulk invite + SSO (coming soon)
Need more than 100 teammates?

Custom contracts add SSO / SAML, custom audit retention, a DPA, and on-call support.

Contact sales
FAQ

Questions, answered.

What is Klavex?

Klavex is a CLI-first secrets manager. Instead of keeping a .env file on disk, the Klavex CLI pulls your environment variables from an encrypted vault and injects them into your app's process at runtime — so the plaintext secrets never sit in a file.

How is Klavex different from Doppler or Infisical?

Same core idea — a central vault instead of scattered .env files — but Klavex is built to be the simplest and cheapest option for solo developers and small teams: one command to import your existing .env, flat per-team pricing, and a deliberately small feature set.

Can AI coding agents read my secrets if I use Klavex?

Coding agents like Cursor, Claude Code, and Copilot can read any file in your project, including .env. Klavex removes the file — secrets live in the process environment at runtime, not on disk — so there's no plaintext file for an agent to open. You can also scope each agent to only the environments you choose.

Where are my secrets stored, and is it secure?

Your secrets are envelope-encrypted in the Klavex vault and only decrypted into your process when you run a command. Access is scoped per environment, so a given token only ever sees the secrets you grant it.

Is Klavex free?

Yes — Klavex is free for solo developers. Teams pay a flat per-team price as they grow, with no per-secret or per-request metering.

Does Klavex work with my existing .env file?

Yes. One command imports an existing .env into Klavex. From there you can delete the file and run your app with the same variables injected at runtime.

Get started

Your secrets don't belong on disk.

Install in 30 seconds. Migrate your first .env in 5 minutes. Sleep better tonight knowing the agents on your machine don't have a free copy of your AWS keys.

$ pip install klavex
$ klavex login
Free Solo plan · No credit card to start · Cancel anytime