How can I set it up without API keys

[TencerDe]

So, I tried setting it up usingg ollama and it ran on my CLI locally but it is asking fir Claude account API, anyone to help a new dev here.

[Nickalus12]

It sounds like the CLI is defaulting to the Anthropic provider. To use Ollama, you need to tell OpenClaude to use the OpenAI-compatible bridge.

The simplest way is to run openclaude and then use the /provider command for the guided setup. If you would rather use environment variables, set these before launching:

export CLAUDE_CODE_USE_OPENAI=1
export OPENAI_BASE_URL=http://localhost:11434/v1
export OPENAI_MODEL=llama3.1:8b

A common gotcha is that the OPENAI_MODEL name must exactly match a model you have already pulled via ollama pull. If it doesn't match, the app might hang or error out during the connection check without a specific message.

[alucard366]

i tried it and the provider doesn't change

[Uffynn]

how did you manage to try? I cant even get to type provider, because it automatically defaults to the Anthropic. Also before even open the app if I type 'openclaude --provider' nothing happens, its an unknown command

[alucard366]

I tried applying the provider before opening with ""$env:CLAUDE_CODE_USE_OPENAI="1"
$env:OPENAI_BASE_URL="https://openrouter.ai/api/v1"
$env:OPENAI_API_KEY="API key"
$env:OPENAI_MODEL="qwen/qwen3.6-plus:free""

and tried with OpenClaude opened with "/provider"

[alucard366]

the command s /provider

[manatactical]

Same issue here. Opening openclaude has anthropic as default. Can't use /provider inside of claude has I'm prompted with selecting the theme and the login method. When trying to change the env, it doesnt recognize it's local and ask for an API key...

[gnanam1990]

Ollama in OpenClaude currently works through the OpenAI-compatible provider path, not a separate native Ollama login/setup flow.

So for local Ollama, you should not need a Claude API key, but OpenClaude does need to start with the correct provider/model from the beginning.

Please retry on the latest main or latest release with one of these:

openclaude --provider ollama --model llama3.1:8b

or:

export CLAUDE_CODE_USE_OPENAI=1
export OPENAI_BASE_URL=http://localhost:11434/v1
export OPENAI_MODEL=llama3.1:8b

Also make sure that exact model is already pulled in Ollama.

If it still opens with Anthropic/login behavior on the latest main, then this is likely more than setup confusion and we should treat it as a startup/provider-selection bug.

If that happens, please share:

• the startup banner
• /status
• whether actual prompts still go through Ollama or not

[Vasanthdev2004]

Yes — Ollama should work without a Claude API key.

The important detail is that in current OpenClaude, Ollama goes through the OpenAI-compatible provider path. If startup is already dropping you into Anthropic/login, set the provider before launch, not after.

Use one of these:

openclaude --provider ollama --model llama3.1:8b

or set env vars before launch:

macOS / Linux
export CLAUDE_CODE_USE_OPENAI=1
export OPENAI_BASE_URL=http://localhost:11434/v1
export OPENAI_MODEL=llama3.1:8b

Windows PowerShell
$env:CLAUDE_CODE_USE_OPENAI=1
$env:OPENAI_BASE_URL=http://localhost:11434/v1
$env:OPENAI_MODEL=llama3.1:8b

Two important clarifications:

• openclaude --provider by itself is not enough; it needs a value such as ollama
• /provider is useful once you are inside the app, but if startup is already jumping into Anthropic/login, configure the provider before launch

If it still lands in Anthropic/login after using the exact command above, please paste:

• how you installed OpenClaude (npm -g package vs source checkout)
• openclaude --version
• the startup banner
• bun run doctor:runtime if you are running from source