Claude Code
One command. The fastest client to get working, and the easiest to debug.
Kind MCP client
Transport Remote HTTP, added from the CLI.
Config —
Setup
- Create a key in the console and export it as UNZOI_KEY.
- Run the add command below.
- Run /mcp inside Claude Code to confirm the server connected and the six tools are listed.
- Ask something that needs current reporting — the tool call appears in the transcript.
claude mcp add --transport http unzoi https://api.unzoi.com/mcp \
--header "x-api-key: $UNZOI_KEY"Or, to share it with a repository as .mcp.json
claude mcp add --transport http --scope project unzoi https://api.unzoi.com/mcp \
--header "x-api-key: ${UNZOI_KEY}"
What differs here
- Scope matters. The default is local to you and this directory; `--scope user` follows you across projects, and `--scope project` writes .mcp.json for the repository — which is the one to use for a team, with the key left as an environment-variable reference rather than a literal.
- `/mcp` inside a session lists connected servers and their tools, which makes this the best client to debug an MCP problem in before trying anywhere else.
- Claude reads tool descriptions carefully and reliably picks list_stories over search_news when a question is about an event rather than about coverage — which is the behaviour these descriptions are written for.
Next
- Get a key at console.unzoi.com, if you have not.
- The tool reference — what the model is reading when it chooses between the six tools.
- Stories for context windows — the one habit that most changes how well an agent works against this API.
- Limits — an agent that loops can spend a monthly quota faster than you expect.
- Claude Code's own MCP documentation — authoritative for anything about the client itself, including config paths, which move.
Endpoint, for copying: https://api.unzoi.com/mcp