Connect Contextick to Claude Code, Codex CLI, or Gemini CLI

Command-line AI tools all connect the same way in principle: you add Contextick's endpoint to the tool's MCP config with an API key. What differs is the file each one reads and the exact field names it expects — so the snippets below aren't interchangeable. Pick your tool.

What you'll need

Two things about the key. It's shown once, right after you create it — copy it then, because you can't view it again. And the name you give the key becomes the origin recorded on everything saved through it, so name it after the tool you're connecting.

The examples below read the key from CONTEXTICK_API_KEY, so set it in your shell first:

export CONTEXTICK_API_KEY="your-api-key"

How the connection works

Contextick is a standard remote MCP server over streamable HTTP. Every tool below needs the same three things from you — the endpoint https://contextick.ai/mcp, an Authorization: Bearer <your API key> header, and a name for the server. Anything else in the snippets is that tool's own spelling of those three things.

Claude Code

One command, no file editing:

claude mcp add --transport http contextick https://contextick.ai/mcp \
  --header "Authorization: Bearer $CONTEXTICK_API_KEY"

If you'd rather write the config by hand, add this to .mcp.json:

{
  "mcpServers": {
    "contextick": {
      "type": "http",
      "url": "https://contextick.ai/mcp",
      "headers": {
        "Authorization": "Bearer ${CONTEXTICK_API_KEY}"
      }
    }
  }
}

"type": "http" is required here. Claude Code reads an entry with no type as a local stdio server, so omitting it is a hard error rather than a silent fallback — you'll see "has a url but no type". This is the one field people copy away from other tools' docs and lose.

Check it with claude mcp list.

Codex CLI

Codex uses TOML, not JSON. Add this to ~/.codex/config.toml (or .codex/config.toml in a project):

[mcp_servers.contextick]
url = "https://contextick.ai/mcp"
bearer_token_env_var = "CONTEXTICK_API_KEY"

Two things to notice: the table is mcp_servers in snake_case, not mcpServers. And Codex takes the name of the environment variable holding your key rather than the key itself — so the key never lands in the file.

The codex mcp add command is for local stdio servers only; remote servers like Contextick are configured by editing the file.

Gemini CLI

Add this to ~/.gemini/settings.json (or .gemini/settings.json in a project):

{
  "mcpServers": {
    "contextick": {
      "httpUrl": "https://contextick.ai/mcp",
      "headers": {
        "Authorization": "Bearer ${CONTEXTICK_API_KEY}"
      }
    }
  }
}

Gemini CLI calls the endpoint field httpUrl. It also accepts url, but that means the older SSE transport — use httpUrl.

Check it worked

Test it in two parts, and make the second part a new session. If you ask in the same one, the tool can answer from what you just said without ever calling Contextick — so a broken connection would still look like it worked.

  1. Save something you'd actually want back. At the end of a session where you worked something out, say: "Save the migration steps we just worked out to Contextick."
  2. Quit, start a new session, and ask: "What are the migration steps we worked out?"

If it comes back in the new session, you're connected — and you've just seen the point of it: you didn't have to explain it again. That note is also available in Claude, ChatGPT, or any other AI you connect to Contextick, and anything you saved from a chat app is readable here too.

If it doesn't show up

Official docs

← All connection guides