The fastest path is the bootstrap installer. It creates a managed local runtime, configures the selected MCP client, and keeps the TypeSafe credential outside client configuration.
- Node.js 20+
- Git
- npm
- a TypeSafe API key / applicable TypeSafe access and credits
- the target AI client
Codex on macOS/Linux:
curl -fsSL https://raw.githubusercontent.com/Afloat16/jev-mcp/v0.5.0/scripts/install.sh | bash -s -- codexCodex on Windows PowerShell:
& ([scriptblock]::Create((irm https://raw.githubusercontent.com/Afloat16/jev-mcp/v0.5.0/scripts/install.ps1))) -Target codexOther target IDs:
claude-code
kimi
zcode
cursor
gemini
windsurf
agents
vscode
Configure every detected user-level client on macOS/Linux:
curl -fsSL https://raw.githubusercontent.com/Afloat16/jev-mcp/v0.5.0/scripts/install.sh | bashThe installer creates:
~/.jev-mcp/runtime # managed local runtime
~/.jev-mcp/.env # local credential file
The API key is entered with hidden input and is not placed in the target client's MCP configuration.
See INSTALLATION.md for all clients and lifecycle commands.
MCP tool discovery normally happens at client/session startup. Restart the client or open a new session after installation.
Use synthetic state and ask the host model for a second opinion on a bounded decision such as:
retry / rollback / escalate
Do not use production secrets or customer data as test input.
deterministic evidence
>
host-model repository-aware reasoning
>
Jev probabilistic advice
Anything placed in Jev state is sent to the configured TypeSafe API
endpoint. Read SECURITY-MODEL.md before using the tool with
sensitive projects.
git clone https://github.com/Afloat16/jev-mcp.git
cd jev-mcp
./setup.shOn Windows PowerShell use ./setup.ps1.