Connect your tools

Connecting a tool lives under Integrations in the dashboard, and there are two ways to do it:

• One-click. Pick the client on the Integrations page and hit install. This works through the local daemon on the target device: the dashboard shows each paired device with its online status, and installs, updates, and uninstalls run on that machine with one click — no terminal needed.
• Copy the command. Every tool also has a one-line installer, listed on the same page and below, that bakes in the gateway URL and your key. Swap sk-llm-YOUR_KEY for your own key from the dashboard.

Either way the result is non-destructive: the installer backs up and merges into your existing config instead of overwriting it, and the one-click uninstall reverses cleanly.

Claude Code

curl -fsSL "https://openllm.sh/api/setup/claude-code/install.sh" \
  | OPENLLM_API_KEY="sk-llm-YOUR_KEY" bash

Merges into ~/.claude/settings.json (preserving your hooks, permissions, and model), sets ANTHROPIC_BASE_URL to your gateway, and maps Opus, Sonnet, and Haiku to ultra, plus, and lite. Needs jq or python3 to merge safely. Restart Claude Code after install.

Codex

curl -fsSL "https://openllm.sh/api/setup/codex/install.sh" \
  | OPENLLM_API_KEY="sk-llm-YOUR_KEY" bash

Adds an openllm provider block to ~/.codex/config.toml and selects it, defaulting the model to the ultra chain. It also fetches your model catalog so Codex sizes auto-compaction correctly and stops warning about missing metadata.

Kimi CLI

curl -fsSL "https://openllm.sh/api/setup/kimi-code/install.sh" \
  | OPENLLM_API_KEY="sk-llm-YOUR_KEY" bash

Adds an openllm provider to ~/.kimi-code/config.toml and writes a model entry for each of your activated models, which the Kimi CLI requires.

Raycast

curl -fsSL "https://openllm.sh/api/setup/raycast/install.sh" \
  | OPENLLM_API_KEY="sk-llm-YOUR_KEY" bash

Merges an openllm custom provider into Raycast's providers.yaml with every activated model and its real context window. Needs a Raycast plan that supports custom AI providers. After install, enable Custom Providers under Settings → AI and restart Raycast.

OpenLLM CLI

curl -fsSL "https://openllm.sh/api/setup/cli/install.sh" | bash

Installs openllmc, a self-contained binary (no node or bun needed) that serves one MCP server exposing the full gateway API, semantic code and docs search, and persistent memory. It verifies its download against a published SHA-256 checksum and self-updates against the gateway's pinned release. After a sandboxed install, run ~/.openllm/bin/openllmc setup to finish shell completion.

The local daemon

Subscription accounts (Claude Code, ChatGPT, Kimi over OAuth) need a small daemon running on your machine, because those credentials never route through the cloud.

curl -fsSL "https://openllm.sh/api/setup/daemon/install.sh" \
  | OPENLLM_API_KEY="sk-llm-YOUR_KEY" bash

Installs openllmd, runs subscription inference locally against the official vendor CLIs, and sets up a launch agent (macOS) or systemd user unit (Linux) so it stays running across logout and reboot. Control it with openllmd start, openllmd stop, and openllmd status. See Providers and Compliance for why this runs locally.

The daemon is also what powers the one-click installs above: once a device is paired, the Integrations page can install, update, or uninstall any client on that machine remotely.