Allow using Gemini CLI for free Gemini requests

[skyfallwastaken]

I'm going to apologise in advance if this is a unreasonable request, but Gemini CLI has a very generous free tier and it would be quite nice to route Gemini requests through it (especially since it costs Google the same amount of money anyway)

[BasixKOR]

Note that the authentication scheme Gemini CLI uses is public: https://github.com/google-gemini/gemini-cli/blob/main/packages/core/src/code_assist/oauth2.ts

[thdxr]

this would honestly be a great PR it might fit in very nicely into how claude pro/max is implemented

[sockthedev]

one tui to rule them all...

[izyuumi]

Does something like this work?

[aryasaatvik]

got it working with with custom ai sdk provider. here is my config ~/.config/opencode.json.

this needs you to use gemini cli to login initially

https://gist.github.com/aryasaatvik/4eb60e53290b4d483bd1fe40dae6fc48

[fer-ri]

any update? I open gist from @aryasaatvik but seems not working anymore too.

[hrstoyanov]

With Qwen Coder CLI now out, this is becoming very interesting:

• Gemini CLI and Qwen CLI are very similar (Qwen CLI is fork of the Gemini cli code)
• Both provide very generous daily limits of free requests! They are practically free for individuals ....
• Both use OAuth for the free personal access, not tokens.

I suspect more (Chineses) AI labs will clone Gemini CLI and couple it with OAuth + generaous free daily usage quota.

That is an opportunity for OpenCode to rule them all, @thdxr

[frap129]

In case anyone else is also trying to get this working, I was able to do it using https://github.com/router-for-me/CLIProxyAPI.

I configured it per the README on that project, and set up the provider in opencode. The only quirk is that I was getting invalid JSON errors when setting it up as an OpenAI compatible provider, it needed to be configured as a google provider:

  "provider": {
    "cli-proxy-api-google": {
      "npm": "@ai-sdk/google",
      "name": "CLIProxyAPI (Google)",
      "options": {
        "baseURL": "http://localhost:8317/v1beta",
        "apiKey": "<key you set in the CLIProxyAPI config>"
      },
      "models": {
        "gemini-2.5-pro": {
          "name": "Gemini 2.5 Pro"
        },
        "gemini-2.5-flash": {
          "name": "Gemini 2.5 Flash"
        },
        "gemini-2.5-flash-lite": {
          "name": "Gemini 2.5 Flash Lite"
        }
      }
    }
  }

Hope this helps

[fer-ri]

In case anyone else is also trying to get this working, I was able to do it using https://github.com/router-for-me/CLIProxyAPI.

I configured it per the README on that project, and set up the provider in opencode. The only quirk is that I was getting invalid JSON errors when setting it up as an OpenAI compatible provider, it needed to be configured as a google provider:

  "provider": {
    "cli-proxy-api-google": {
      "npm": "@ai-sdk/google",
      "name": "CLIProxyAPI (Google)",
      "options": {
        "baseURL": "http://localhost:8317/v1beta",
        "apiKey": "<key you set in the CLIProxyAPI config>"
      },
      "models": {
        "gemini-2.5-pro": {
          "name": "Gemini 2.5 Pro"
        },
        "gemini-2.5-flash": {
          "name": "Gemini 2.5 Flash"
        },
        "gemini-2.5-flash-lite": {
          "name": "Gemini 2.5 Flash Lite"
        }
      }
    }
  }

Hope this helps

Thanks a lot

[hrstoyanov]

So CLIProxyAPI doesn't not use OAuth, you need to provide API key?

[frap129]

@hrstoyanov It uses oauth. The API key in the opencode config is for connecting to the local CLIProxyAPI server.

[nitish-raj]

@frap129 is this still working for you? For me it doesn't seem to connect. Could you share your config file rom CLIProxyAPI?

[frap129]

@nitish-raj Yeah, sometimes messages hang and I have to abort and resend, but otherwise it works well. My config is just the example from the CLIProxyAPI readme, with different local keys, a different auth directory, and no API keys added:

# Server port
port: 8317
# Management API settings
remote-management:
  # Whether to allow remote (non-localhost) management access.
  # When false, only localhost can access management endpoints (a key is still required).
  allow-remote: false
  # Management key. If a plaintext value is provided here, it will be hashed on startup.
  # All management requests (even from localhost) require this key.
  # Leave empty to disable the Management API entirely (404 for all /v0/management routes).
  secret-key: "REDACTED"
# Authentication directory (supports ~ for home directory)
auth-dir: "~/.cache/cli-proxy-api"
# Enable debug logging
debug: true
# Proxy URL. Supports socks5/http/https protocols. Example: socks5://user:pass@192.168.1.1:1080/
proxy-url: ""
# Number of times to retry a request. Retries will occur if the HTTP response code is 403, 408, 500, 502, 503, or 504.
request-retry: 3
# Quota exceeded behavior
quota-exceeded:
  switch-project: true # Whether to automatically switch to another project when a quota is exceeded
  switch-preview-model: true # Whether to automatically switch to a preview model when a quota is exceeded
# API keys for authentication
api-keys:
  - "REDACTED"

[JunarisAlf]

@frap129 very nice, the openAI compatible is work well on kilo code, but i'm wondering why is not on opencode