End-to-end encrypted secrets for teams, machines, and AI agents.

Secrets are encrypted in your browser or CLI before they ever leave the machine, so seekrit stores ciphertext and nothing else — a full database dump reveals no secret values. One command injects the right secrets everywhere your code runs: local dev, Docker, CI, Kubernetes, and agent sandboxes.

Anything an agent can read, an agent can be talked into handing over. So don't give it the key — give it a placeholder, and broker the real credential at the network boundary:

npx -y @seekrit/cli proxy run --preset openai

Your workload sends Authorization: Bearer {{seekrit:OPENAI_API_KEY}}. The proxy substitutes the real, decrypted value on the way out, against a default-deny allowlist that bounds which hosts, methods, and paths each secret may reach, and audits every substitution by name. The agent never holds a credential, so there is nothing there to leak.

Unlike a hosted credential broker, the proxy runs on your machine and seekrit holds only ciphertext — the service cannot read your secrets, and cannot rewrite your egress policy either. Policy bundles are signed in your browser and verified against signers pinned in the proxy's own config.

Most of these are read-only mirrors published from seekrit's monorepo, so the code that holds your token and decrypts plaintext is auditable before you trust it. Issues and PRs are welcome; accepted changes get ported upstream.

Agents & MCP

• agent-plugin — both seekrit MCP servers plus the skills that keep API keys out of your agent's transcript. One install.
• openclaw-plugin — seekrit as an OpenClaw SecretRef provider, so the gateway config names credentials instead of storing them.

SDKs — resolve and decrypt with a service token

• js-sdk — JavaScript / TypeScript. WebCrypto + fetch, so it runs on Node, Bun, Deno, browsers and Cloudflare Workers.
• python-sdk — Python, plus httpx transport and LangChain / Pydantic AI adapters.
• go-sdk — Go.
• ruby-sdk — Ruby.

Infrastructure & CI

• terraform-provider-seekrit — apps, environments, groups, tokens, grants and secrets as code. Write-only arguments and ephemeral resources keep values out of Terraform state.
• github-action — load secrets into a GitHub Actions job, decrypted on the runner and masked in the logs.
• helm-charts — including seekrit-eso, for Kubernetes.
• seekrit-sdk-server — cluster-side resolver that syncs secrets into Kubernetes through stock External Secrets Operator, decrypting inside your cluster.

• Site & docs — seekrit.dev · Quickstart · Agent egress proxy
• Console — app.seekrit.dev
• Security — the threat model. Responsible disclosure: security@seekrit.dev