Code

Closed beta: lightning code is in closed beta, and we're rolling it out to more people. If you'd like to take part, contact support@lightning.ai.

code.lightning.ai serves models for coding agents through an OpenAI-compatible API at https://code.lightning.ai/v1. The lightning code commands set up a coding tool to use it, or create an API key for any other tool. Requests bill the organization you choose, which needs a Pro, Teams or Enterprise plan.

The commands set up these tools:

Tool	Name for the commands
OpenCode	`opencode`
pi	`pi`
Codex	`codex`
DeepSeek Harness	`dsh`
Cursor	`cursor`

The same steps, with your organization filled in, are under Settings > Lightning Code, in a plan's Coding tools section.

Before you begin

The Lightning CLI, signed in. Install or update it with pip install -U lightning-sdk and run lightning login; see CLI. Where pip is missing or refuses to install outside a virtual environment, as on recent Ubuntu releases, use uv tool install -U lightning-sdk instead. On a machine without a browser, set LIGHTNING_USER_ID and LIGHTNING_API_KEY instead of running lightning login.
An organization on a Pro, Teams or Enterprise plan. Compare plans on the pricing page.
The coding tool installed. Each tool's section below shows how. pi and DeepSeek Harness need Node.js 22.19 or newer.

Set up a tool

This adds Lightning as a model provider in the tool and creates an API key that bills your organization.

Run the setup with the tool's name, replacing my-org with your organization's name:

lightning code setup opencode --org my-org

Without --org, the command lists your organizations with their plans and asks which one pays. In scripts and other non-interactive shells, --org is required.

The output names the files it changed:

OpenCode now uses code.lightning.ai, billed to My Org [my-org].
  Provider: lightning in ~/.config/opencode/opencode.json
  Key file: ~/.local/share/opencode/auth.json
  API key:  opencode on my-laptop 2026-10-09
  Default:  lightning/deepseek-v4.1-flash

Start the tool. Its section below says how.

For every tool, the setup:

Adds a provider called lightning and leaves the rest of the tool's config as it is, comments included. Before changing a config file you already had, it backs it up by appending .lightning-backup to its filename. Files that hold keys aren't copied.
Saves the API key where the tool keeps its own keys, not in a file you'd share or commit.
Makes DeepSeek V4.1 Flash the default model only if you don't have one yet.
Takes the model list from https://code.lightning.ai/v1/models, so run the setup again to pick up new models.

Setup options

Option	What it does
`--org`	Organization that pays, by name.
`--model`	Makes this model the default, for example `glm-5.3`.
`--dry-run`	Shows the changes to the config without writing anything or creating a key.
`--rotate-key`	Creates a new API key and revokes the old one.
`--force`	Replaces a Lightning setup that you added by hand.

To bill a different organization, run the setup again with its name. This creates a key there and revokes the old one.

OpenCode

curl -fsSL https://opencode.ai/install | bash
lightning code setup opencode --org my-org
opencode

The provider goes into ~/.config/opencode/opencode.json, or opencode.jsonc if that's the file you have. The key goes into OpenCode's ~/.local/share/opencode/auth.json. If your config lists providers in enabled_providers, the setup adds lightning to it. Switch models with /models.

The installer adds opencode to your PATH in your shell's startup file, so start it from a new terminal.

pi

curl -fsSL https://pi.dev/install.sh | sh
lightning code setup pi --org my-org
pi

The provider goes into ~/.pi/agent/models.json, the key into pi's auth.json and the default model into settings.json. When PI_CODING_AGENT_DIR is set, all three are in that directory instead. Switch models with /model.

The installer offers to install Node.js if yours is missing or too old. When it offers to start pi, say no: pi has no models until you run the setup. Start pi from a new terminal, so it's on your PATH.

Codex

npm install -g @openai/codex
lightning code setup codex --org my-org
codex --profile lightning

Codex gets a profile of its own, ~/.codex/lightning.config.toml, which it uses on top of your config.toml when you start it with --profile lightning. Your usual Codex setup doesn't change. The profile holds the key, so only you can read it. Pick another model with -m, for example codex --profile lightning -m lightning-ai/glm-5.3.

Image input through Codex isn't supported yet. The other tools send images.

DeepSeek Harness

DeepSeek Harness needs Node.js 22.19 or newer. npm installs it on an older Node.js without a warning, but then dsh exits without doing anything. Check with node --version. In a Studio, run nvm install 22 first.

npm install -g @deepseek-ai/dsh
lightning code setup dsh --org my-org
dsh web

The provider and default model go into the web and headless profiles, in ~/.dsh/profiles/<profile>/cordis.patch.yml, next to your other entries. The key goes into ~/.dsh/.credentials.yaml as LIGHTNING_CODE_API_KEY.

dsh web prints a local address with a token, such as http://127.0.0.1:3080/?token=.... Open it in your browser and pick a Lightning model under Model in the message box.

Cursor

Cursor's settings can't be changed from outside Cursor, so the setup creates a key and prints the steps to add it in Cursor. You need a paid Cursor plan.

lightning code setup cursor --org my-org

Cursor now uses code.lightning.ai, billed to My Org [my-org].
  API key: cursor on my-laptop 2026-10-09

Finish in Cursor (needs a paid Cursor plan):
  1. Open Cursor Settings (Cmd+Shift+J on macOS, Ctrl+Shift+J elsewhere) and go to Models.
  2. Under API Keys, paste this into OpenAI API Key and press Enter:
       sk-lit-...
  3. Turn on "Use OpenAI API Key".
  4. Turn on "Override OpenAI Base URL" and enter:
       https://code.lightning.ai/v1
  5. At the top of Models, add these model names and make sure they're switched on:
       lightning-ai/glm-5.3
       lightning-ai/glm-5.3-flash
       lightning-ai/deepseek-v4.1-flash
  6. Open a new chat and pick lightning-ai/deepseek-v4.1-flash.

The key is shown only this once. While the base URL override is on, Cursor sends all OpenAI models to Lightning, so its own GPT and Auto models stop working until you turn it off. Tab completion keeps using Cursor's models, and the cursor-agent CLI can't use Lightning models.

Check the setup

lightning code status

OpenCode: billed to my-org
  API key: opencode on my-laptop 2026-10-09 (01jexamplekey0000000000000)
  Config:  ~/.config/opencode/opencode.json
  Default: lightning/deepseek-v4.1-flash
pi: not set up
Codex: billed to my-org
  ...

Add --json for machine-readable output.

Remove the setup

lightning code remove opencode

This removes the lightning provider, a default model that uses it, and the API key, which it also revokes. Files the setup created are deleted, and the rest of your config stays as it is. To leave the key active, add --keep-key. For Cursor, the command revokes the key and prints how to turn the override off.

Use other tools

For tools that speak the OpenAI API, create a key with lightning code token and point the tool at https://code.lightning.ai/v1:

export OPENAI_BASE_URL=https://code.lightning.ai/v1
export OPENAI_API_KEY=$(lightning code token --org my-org)

The key is printed once, so store it straight away. Name it with --name, or add --json to get the key with its ID, organization and base URL.

Requests name the model by its full ID:

Model	ID	Context window	Image input
GLM-5.3	`lightning-ai/glm-5.3`	1,048,576	No
GLM-5.3 Flash	`lightning-ai/glm-5.3-flash`	253,952	Yes
DeepSeek V4.1 Flash	`lightning-ai/deepseek-v4.1-flash`	253,952	Yes

For the current list, with each model's output limit and supported parameters, request https://code.lightning.ai/v1/models. It doesn't need a key.

Keys from lightning code are listed and deleted with lightning api-key.

Troubleshooting

... is on the Free plan, which doesn't include code.lightning.ai. The organization needs a Pro, Teams or Enterprise plan. Upgrade it with the link in the message, or pass another organization to --org.

Coding tools are turned off for ... An organization admin has turned off coding tools. Ask them to turn them back on, or pass another organization to --org.

... already has a Lightning setup that lightning code didn't make. You added one by hand, for example from another guide. Run the setup again with --force to replace it. Config files are backed up first.

No such command 'code', or Invalid value for 'TOOL'. Your Lightning CLI is too old. Update it with pip install -U lightning-sdk.