Unable to select Claude Sonnet as model for Copilot provider

[anderssv]

From goose configure, choose Copilot:

┌   goose-configure 
│
◇  What would you like to configure?
│  Configure Providers 
│
◇  Which model provider should we use?
│  Github Copilot 
│
◇  Enter a model from that provider:
│  claude-3.7-sonnet
│
◇  Server error: Object {"error": String("An internal server error occurred.")}
│
└  Failed to configure provider: init chat completion request with tool did not succeed.

This is different from failing with a bad model string. We are using sonnet in VSCode and IDEA at the org level. The default (gpt-4o) works fine.

[kcalixto]

i'm also having the same issue!

[matthewgrossman]

Also having this issue for claude-3.7-sonnet.

It's strangely difficult to find the identifiers for the available models (and it differs per-account), but taking hints from aider's docs, you can list all the models via:

apikey=$(jq -r 'first(.[]) | .oauth_token' ~/.config/github-copilot/apps.json)
curl -s -H "Authorization: Bearer $apikey" https://api.githubcopilot.com/models | jq -r '.data[].id'

Models that have worked:

• gpt-4o
• o3-mini

Models that seem broken:

• claude-3.5-sonnet
• claude-3.7-sonnet
• gpt-4.1

gpt-4.1 gives:

◇  Enter a model from that provider:
│  gpt-4.1
│
◇  Request failed: error decoding response body
│
└  Failed to configure provider: init chat completion request with tool did not succeed.

While, the claude-* models give:

◇  Enter a model from that provider:
│  claude-3.7-sonnet
│
◇  Server error: Object {"error": String("An internal server error occurred.")}
│
└  Failed to configure provider: init chat completion request with tool did not succeed.

[fxck]

Welcome to goose! Let's get you set up with a provider.
  you can rerun this command later to update your configuration

┌   goose-configure 
│
◇  Which model provider should we use?
│  Anthropic 
│
●  ANTHROPIC_API_KEY is set via environment variable
│  
◇  Would you like to save this value to your keyring?
│  No 
│
●  ANTHROPIC_HOST is set via environment variable
│  
◇  Would you like to save this value to your keyring?
│  No 
│
◇  Model fetch complete
│
◇  Enter a model from that provider:
│  claude-3.5-sonnet
│
◇  Request failed: Request failed with status: 404 Not Found
│
└  Failed to configure provider: init chat completion request with tool did not succeed.

  Warning: We did not save your config, inspect your credentials
   and run 'goose configure' again to ensure goose can connect

[gugahoi]

Just wanted to report that I was able to set Claude as the model by editing the config file directly:

❯ goose session
starting session | provider: github_copilot model: claude-3.7-sonnet
    logging to /Users/guga/.local/share/goose/sessions/20250523_101257.jsonl
    working directory: /Users/guga/

Goose is running! Enter your instructions, or try asking what goose can do.

( O)> hello
# Hello!

Welcome to Goose! I'm here to help you with various tasks. I have access to several tools that allow me to help you work with files, run commands, and more.

How can I assist you today? I can help with:

- Writing, editing, or viewing code files
- Running shell commands
- Finding files or information within your project
- Working with the Flux project you have open
- Capturing screenshots if needed

config.yaml

GOOSE_MODEL: "claude-3.7-sonnet"
extensions:
  developer:
    bundled: true
    display_name: Developer
    enabled: true
    name: developer
    timeout: 300
    type: builtin
GOOSE_PROVIDER: github_copilot

[taniacryptid]

Thank you for sharing what worked for you, @gugahoi ! Going to try to get more eyes on this.

[michaelneale]

interesting - the copilot provider is the common thread here (interesting that it works with default). Not sure how to test if there is a way to get test creds to try things out. When it saves the configure, it tries to do a test call (to check tools and so on work) and that is what is failing, but obviously nothing else as it seems to work

[wonrax]

Setting the Anthropic models in config file does not work for me though. The chats still throw the internal server error occurred.

[kcalixto]

I'm having a similar issue, but not even gpt-4o is working. My copilot access was grant through SSH with org SSO authorization, and works just fine with my neovim "github/copilot.vim" plugin, also works in vs code

nvim❯ goose update
Downloading stable release: goose-x86_64-apple-darwin.tar.bz2...
Extracting goose-x86_64-apple-darwin.tar.bz2 to temporary directory...
Moving goose to /Users/kcalixto/.local/bin/goose
Skipping 'goose configure', you may need to run this manually later

nvim❯ goose configure
┌   goose-configure
│
◇  What would you like to configure?
│  Configure Providers
│
◇  Which model provider should we use?
│  Github Copilot
│
◇  Model fetch complete
│
◇  Enter a model from that provider:
│  gpt-4o
│
◇  Execution error: failed to get api info after 3 attempts
│
└  Failed to configure provider: init chat completion request with tool did not succeed.

nvim❯ gh auth status
github.com
  ✓ Logged in to github.com account kcalixto (keyring)
  - Active account: true
  - Git operations protocol: ssh
  - Token: ************************************
  - Token scopes: 'admin:public_key', 'gist', 'read:org', 'repo'

nvim❯ gh copilot suggest "how to print hello world"

Welcome to GitHub Copilot in the CLI!
version 1.1.0 (2025-02-10)

I'm powered by AI, so surprises and mistakes are possible. Make sure to verify any generated code or suggestions, and share feedback so that we can learn and improve. For more information, see https://gh.io/gh-copilot-transparency

? What kind of command can I help you with?
> generic shell command

Suggestion:

  echo "Hello, World!"

? Select an option
> Exit

i've also tried @gugahoi solution:

nvim❯ goose
starting session | provider: github_copilot model: claude-3.7-sonnet
    logging to /Users/kcalixto/.local/share/goose/sessions/20250523_115129.jsonl
    working directory: /Users/kcalixto/.config/nvim

Goose is running! Enter your instructions, or try asking what goose can do.

( O)> hello
  2025-05-23T14:51:32.310504Z  WARN goose::providers::githubcopilot: failed to refresh api info: HTTP status client error (404 Not Found) for url (https://api.github.com/copilot_internal/v2/token)
    at crates/goose/src/providers/githubcopilot.rs:174

  2025-05-23T14:51:32.470390Z  WARN goose::providers::githubcopilot: failed to refresh api info: HTTP status client error (404 Not Found) for url (https://api.github.com/copilot_internal/v2/token)
    at crates/goose/src/providers/githubcopilot.rs:174

  2025-05-23T14:51:32.648733Z  WARN goose::providers::githubcopilot: failed to refresh api info: HTTP status client error (404 Not Found) for url (https://api.github.com/copilot_internal/v2/token)
    at crates/goose/src/providers/githubcopilot.rs:174

Session ended with error: Execution error: failed to get api info after 3 attempts

nvim❯

[wonrax]

@kcalixto I think you need to perform device code authorization using goose configure since it uses Copilot token instead of the gh CLI. Simply setting GOOSE_PROVIDER does not work because it does not acquire and store the token used for requests. The fix is to remove the key GOOSE_PROVIDER in config file and run goose configure to let GitHub Copilot grant goose access. I haven't read much into the codebase so I might be wrong though.

[lbergnehr]

I think the errors have to do with tool usage. I can -- if I configure it manually in the config file -- tell e.g. claude 4 sonnet through github copilot to write me a spring poem and I get an answer. However, when I tell it to write the spring poem to a file, it crashes with the internal server error.

[brianhuster]

Same here, I also think it is a tool usage problem. I have configured Goose to use Claude in config.yaml, and it works as long as it doesn't call my custom MCP server (I have tested that MCP server using MCP inspector, and I don't think there is anything wrong with it)

GPT 4o and o3-mini work fine, but they are stupid when it comes to tool usage

[gugahoi]

I've had issues getting Copilot to interpret images where Gemini seemed to work fine. Both Claude and GPT-4o didn't work for me: https://discord.com/channels/1287729918100246654/1375329182627926196

[Linuus]

Just wanted to report that I was able to set Claude as the model by editing the config file directly:

Thank you! Quoting seems to do the trick 🙌

Or no, it works to configure but still when it's used I get

Ran into this error: Server error: Object {"error": String("An internal server error occurred.")}.

As mentioned above it seems related to when using tools. If asked to write a poem or something it can deliver without errors.

[gugahoi]

I've created a PR to properly list the models from Copilot during goose configure, instead of needing to guess the correct values (#2717). I've obviously not tested all of these and I didn't even realize there were so many available to be honest.

I also saw that in another tool it was reported that claude-3.7-sonnet through Copilot is not as good. This reflects my experience and it's fairly disappointing, if true: yetone/avante.nvim#2093 (comment)