Connect your tools¶

Every tool uses the same address, https://gateway.keera.ch/api. Give each machine and tool a key of its own, so you can replace one without breaking the others.

Claude Code¶

Two commands set up Claude Code on this machine:

keera login --url https://gateway.keera.ch
keera connect claude-code

This gives the machine its own key and writes it into ~/.claude/settings.json, together with every model you may use. As a member, it activates the key an administrator issued for you. The rest of the file stays as it was. Run claude in your project, and /model switches between your models.

• --model <alias> picks the model Claude Code starts with.
• --print shows the settings instead of writing them, for another machine.
• Run the command again to replace the key with a new one.

On a self-hosted gateway, a Claude Team or Enterprise plan can pay instead; see Claude plans.

Codex¶

Set your key as an environment variable, then add the gateway to ~/.codex/config.toml:

model = "keera-speed"
model_provider = "keera"

[model_providers.keera]
name = "Keera"
base_url = "https://gateway.keera.ch/api/v1"
env_key = "KEERA_API_KEY"
wire_api = "responses"

Put the first two lines above the first [section] of the file. Then run codex in your project.

OpenCode, Pi and others¶

keera connect prints the settings for every tool your gateway knows:

keera connect               # the list of tools
keera connect opencode      # the settings for OpenCode
keera connect pi --key laptop

The settings go to the screen and the instructions go to the error output, so you can write the settings straight into a file:

keera connect opencode > ~/.config/opencode/opencode.json

Any OpenAI client¶

Anything that takes an OpenAI base URL and an API key works. Use:

• Base URL: https://gateway.keera.ch/api/v1
• API key: your Keera key, keera_sk_…
• Model: an alias you may use, such as keera-speed

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://gateway.keera.ch/api/v1",
    api_key=os.environ["KEERA_API_KEY"],
)
reply = client.chat.completions.create(
    model="keera-speed",
    messages=[{"role": "user", "content": "Hello"}],
)
print(reply.choices[0].message.content)

curl https://gateway.keera.ch/api/v1/chat/completions \
  -H "Authorization: Bearer $KEERA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "keera-speed", "messages": [{"role": "user", "content": "Hello"}]}'

import os
from anthropic import Anthropic

client = Anthropic(
    base_url="https://gateway.keera.ch/api",
    auth_token=os.environ["KEERA_API_KEY"],
)
reply = client.messages.create(
    model="keera-speed",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello"}],
)
print(reply.content[0].text)

To see which models your key may use:

curl https://gateway.keera.ch/api/v1/models -H "Authorization: Bearer $KEERA_API_KEY"

MCP servers¶

Your administrator can put MCP servers behind the gateway too. Use the same key. keera mcp connect <server> prints the settings, for example:

claude mcp add --transport http github https://gateway.keera.ch/api/mcp/github \
  --header "Authorization: Bearer $KEERA_API_KEY"