Does the agent ever see my key?
No. It only sees the key's name and whether the write succeeded.
@offprompt
Basic information
MCP server that lets Claude Code, Cursor or Codex ask you for API keys on a local page and write them to .env. The agent only ever sees key names.



Add this server to your MCP-compatible client using the configuration below.
{
"mcpServers": {
"offprompt": {
"command": "npx",
"args": [
"-y",
"offprompt",
"mcp"
]
}
}
}Tool list not available yet
This listing does not currently have a tool list to display. This does not mean the server has no tools; check its documentation for supported capabilities.
Fetch the live tool list by running this server in a temporary sandbox using the button above.
offprompt is an open-source (MIT) MCP server that keeps API keys out of your AI agent's context. When Claude Code, Cursor, Codex or any MCP client needs a secret, it asks offprompt. A page opens in your browser on 127.0.0.1, you type or paste the value there, and offprompt writes it straight into .env.local. The transcript, the model's context and the provider's logs only ever see the key's name.
Run npx offprompt init in your project (Node 22 or later). It detects your agents and adds the MCP server for them. To add it by hand, use:
{"mcpServers":{"offprompt":{"command":"npx","args":["-y","offprompt","mcp"]}}}
Then ask your agent to do something that needs a key, for example "add Stripe webhooks". It calls collect_secret, the page opens, and you fill it in.
collect_secret, await_secret and cancel_secret..env with several keys in one go.No. It only sees the key's name and whether the write succeeded.
In your project's `.env.local` (or the file you choose), the same place you'd put them by hand.
Any MCP client, including Claude Code, Cursor and Codex. Try it in the [playground](https://offprompt.dev/playground) or read the [docs](https://docs.offprompt.dev).

Local-first credential broker for MCP. AI coding agents (Claude Code, Cursor, Codex…) call APIs, SSH and databases through scoped presets; secrets stay in your OS keychain and never enter the agent's context.


Word search generator, crossword generator, and sudoku generator + solver as a local-first MCP server. 15 deterministic tools: printable PDF puzzle worksheets, themed word banks, and verifiable LLM evals. From the makers
CronWatch records every run of your scheduled jobs in your own database, and speaks up when one is missed, fails, gets stuck, runs slow or costs too much. CronWatch is free and open-source.

Catch intent conflicts before code conflicts. The open-source coordination protocol for coding agents, built above Git.
Comments