credential-isolation

Give your AI coding agents their own machines — not your credentials. Isolated Linux workstations with full autonomy, on Linux and macOS.

Run Claude Code, Codex, and other command-line AI tools with different providers concurrently — without switching shared configuration.

Per-agent scoped MCP tool proxy — credential isolation, resource scoping, and audit logging for AI agent deployments

Open-source agent middleware — shared memory, credential isolation, and schema learning for AI agents. What one agent learns, every agent knows.

MCP security patterns — 10 production implementations mapped to OWASP MCP Top 10. Prompt injection, credential isolation, HITL governance, session recording. Working attack demos included.

Use-time file-projected tokens, an env-exclusion threat model for the code-writing AI as an insider, per-operation URL splicing, a pure fail-closed tenant×repo routing matrix, and a pre-push target assertion enforced outside the AI sandbox.

Credential-isolated, installable reference deployment of Shutter’s Concorde-based Keyper-status workflow.

Hermetic GitHub identity workspaces for Windows, powered by WSL2