bad request: invalid token: unknown stamp "prod-sdc-01"

[PrimaMateria]

I am getting recently the following error.

Error  03:32:35 PM notify.error CodeCompanion [ERROR] 2025-12-04 15:32:35
Error: bad request: invalid token: unknown stamp "prod-sdc-01"
Error  03:32:35 PM notify.error CodeCompanion [ERROR] 2025-12-04 15:32:35
[chat::_submit_http] Error: {
  body = 'bad request: invalid token: unknown stamp "prod-sdc-01"',
  exit = 0,
  headers = { "content-security-policy: default-src 'none'; sandbox", "content-type: text/plain; charset=utf-8", "strict-transport-security: max-age=31536000", "x-content-type-options: nosniff", "date: Thu, 04 Dec 2025 14:32:35 GMT", "content-length: 56", "x-github-backend: Kubernetes", "x-github-request-id: FBFC:2FCAB:17DDE89:1AE3DCB:69319B83" },
  status = 400
}

The chat was before working. The issue started few days ago without me touching any configuration.
The inline completion are still working, only the main chat is broken.

My current config:

# vim: ft=lua 
{ extraPackages, pkgs, }: ''

  local gheUrl = "https://finapi.ghe.com"

  -- copilot-lua
  -- Run :Copilot auth
  require("copilot").setup({
    nes = { enabled = false }, -- TODO: try later
    auth_provider_url = gheUrl,
    -- nixpkgs node
    copilot_node_command = "${pkgs.nodejs_22}/bin/node",

    suggestion = { enabled = false },
    panel = { enabled = false },
    filetypes = {
      markdown = true,
      help = true,
    },
  })


  -- codecompanion
  -- Available models:
  -- "claude-sonnet-4"
  -- "gemini-2.5-pro"
  -- "grok-code-fast-1"
  -- "claude-3.5-sonnet"
  -- "gpt-5-codex"
  -- "gpt-4o"
  -- "gpt-5"
  -- "claude-sonnet-4.5"
  -- "gpt-4.1"
  -- "gpt-5-mini"

  require("codecompanion").setup({
    opts = {
      log_level = "TRACE",
    },
    adapters = {
      http = {
        copilot_enterprise = function()
          local adapter = require("codecompanion.adapters.http.copilot_enterprise")
          adapter.opts.provider_url = gheUrl
          return adapter
        end,
      },
    },
    strategies = {
      chat = {
        adapter = {
          name = "copilot_enterprise",
          model = "gpt-5",
        },
      },
      inline = {
        adapter = "copilot_enterprise",
        model = "gpt-4.1",
      },
      cmd = {
        adapter = "copilot_enterprise",
        model = "gpt-4.1",
      },
    },
  })


  -- CodeCompanion keymaps
  vim.keymap.set("n", "<leader>ac", "<cmd>CodeCompanionChat Toggle<cr>", { desc = "Toggle CodeCompanion Chat" })
  vim.keymap.set("v", "<leader>ac", "<cmd>CodeCompanionChat Toggle<cr>", { desc = "Toggle CodeCompanion Chat" })
  vim.keymap.set("n", "<leader>aa", "<cmd>CodeCompanionActions<cr>", { desc = "CodeCompanion Actions" })
  vim.keymap.set("v", "<leader>aa", "<cmd>CodeCompanionActions<cr>", { desc = "CodeCompanion Actions" })
  vim.keymap.set("n", "<leader>ai", "<cmd>CodeCompanion<cr>", { desc = "CodeCompanion Inline" })
  vim.keymap.set("v", "<leader>ai", "<cmd>CodeCompanion<cr>", { desc = "CodeCompanion Inline" })
  vim.keymap.set("n", "<leader>an", "<cmd>CodeCompanionChat<cr>", { desc = "New CodeCompanion Chat" })
  vim.keymap.set("n", "<leader>ar", "<cmd>CodeCompanionChat Add<cr>", { desc = "Add to CodeCompanion Chat" })
  vim.keymap.set("v", "<leader>ar", "<cmd>CodeCompanionChat Add<cr>", { desc = "Add selection to CodeCompanion Chat" })
''

Adapter inspection:

03:44:31 PM msg_show.lua_print   :lua print(vim.inspect(require("codecompanion.adapters.http.copilot_enterprise"))) {
  env = {
    api_key = <function 1>
  },
  features = {
    text = true,
    tokens = true
  },
  formatted_name = "Copilot Enterprise",
  get_copilot_stats = <function 2>,
  handlers = {
    chat_output = <function 3>,
    form_messages = <function 4>,
    form_parameters = <function 5>,
    form_tools = <function 6>,
    inline_output = <function 7>,
    on_exit = <function 8>,
    setup = <function 9>,
    tokens = <function 10>,
    tools = {
      format_tool_calls = <function 11>,
      output_response = <function 12>
    }
  },
  headers = {
    Authorization = "Bearer ${api_key}",
    ["Content-Type"] = "application/json",
    ["Copilot-Integration-Id"] = "vscode-chat",
    ["Editor-Version"] = "Neovim/0.12.0"
  },
  name = "copilot_enterprise",
  opts = {
    provider_url = "https://finapi.ghe.com",
    stream = true,
    tools = true,
    vision = true
  },
  roles = {
    llm = "assistant",
    tool = "tool",
    user = "user"
  },
  schema = {
    max_tokens = {
      default = 16384,
      desc = "The maximum number of tokens to generate in the chat completion. The total length of input tokens and generated tokens is limited by the model's context length.",
      mapping = "parameters",
      order = 4,
      type = "integer"
    },
    model = {
      choices = <function 13>,
      default = "gpt-4.1",
      desc = "ID of the model in use. Model availability may depend on you GitHub Enterprise plan.",
      mapping = "parameters",
      order = 1,
      type = "enum"
    },
    n = {
      condition = <function 14>,
      default = 1,
      desc = "How many chat completions to generate for each prompt.",
      mapping = "parameters",
      order = 6,
      type = "number"
    },
    temperature = {
      condition = <function 15>,
      default = 0.1,
      desc = "What sampling temperature to use, between 0 and 2. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic.",
      mapping = "parameters",
      order = 3,
      type = "number"
    },
    top_p = {
      condition = <function 16>,
      default = 1,
      desc = "An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the tokens with top_p probability mass.",
      mapping = "parameters",
      order = 5,
      type = "number"
    }
  },
  show_copilot_stats = <function 17>,
  url = "https://api.githubcopilot.com/chat/completions"
}

CodeCompanion debug window:

-- Adapter: "Copilot Enterprise"
-- Buffer Number: 6
-- Following Buffer: "" (1)

local settings = {
  model = "gpt-5",  -- "gpt-5.1", "claude-haiku-4.5", "gpt-5.1-codex", "gemini-2.5-pro", "gpt-4.1", "gpt-5-mini", "gpt-5.1-codex-mini", "gemini-3-pro-preview", "claude-sonnet-4.5", "gpt-5-codex", "grok-code-fast-1", "gpt-4o", "claude-sonnet-4", "claude-opus-4.5", 
  temperature = 0.1,
  max_tokens = 16384,
  top_p = 1,
  n = 1,
}

local messages = 
{ {
    _meta = {
      cycle = 1,
      id = 59103469,
      index = 1,
      tag = "system_prompt_from_config"
    },
    content = "You are an AI programming assistant named \"CodeCompanion\", working within the Neovim text editor.\n\nYou can answer general programming questions and perform the following tasks:\n* Answer general programming questions.\n* Explain how the code in a Neovim buffer works.\n* Review the selected code from a Neovim buffer.\n* Generate unit tests for the selected code.\n* Propose fixes for problems in the selected code.\n* Scaffold code for a new workspace.\n* Find relevant code to the user's query.\n* Propose fixes for test failures.\n* Answer questions about Neovim.\n* Prefer vim.api* methods where possible.\n\nFollow the user's requirements carefully and to the letter.\nUse the context and attachments the user provides.\nKeep your answers short and impersonal, especially if the user's context is outside your core tasks.\nUse Markdown formatting in your answers.\nDO NOT use H1 or H2 headers in your response.\nWhen suggesting code changes or new content, use Markdown code blocks.\nTo start a code block, use 4 backticks.\nAfter the backticks, add the programming language name as the language ID.\nTo close a code block, use 4 backticks on a new line.\nIf the code modifies an existing file or should be placed at a specific location, add a line comment with 'filepath:' and the file path.\nIf you want the user to decide where to place the code, do not add the file path comment.\nIn the code block, use a line comment with '...existing code...' to indicate code that is already present in the file.\nCode block example:\n````languageId\n// filepath: /path/to/file\n// ...existing code...\n{ changed code }\n// ...existing code...\n{ changed code }\n// ...existing code...\n````\nEnsure line comments use the correct syntax for the programming language (e.g. \"#\" for Python, \"--\" for Lua).\nFor code blocks use four backticks to start and end.\nAvoid wrapping the whole response in triple backticks.\nDo not include diff formatting unless explicitly asked.\nDo not include line numbers in code blocks.\n\nWhen given a task:\n1. Think step-by-step and, unless the user requests otherwise or the task is very simple. For complex architectural changes, describe your plan in pseudocode first.\n2. When outputting code blocks, ensure only relevant code is included, avoiding any repeating or unrelated code.\n3. End your response with a short suggestion for the next user turn that directly supports continuing the conversation.\n\nAdditional context:\nAll non-code text responses must be written in the English language.\nThe current date is 2025-12-04.\nThe user's Neovim version is 0.12.0.\nThe user is working on a Linux machine. Please respond with system specific commands if applicable.\n",
    opts = {
      visible = false
    },
    role = "system"
  }, {
    _meta = {
      cycle = 1,
      id = 1195041320,
      index = 2,
      sent = true
    },
    content = "test",
    opts = {
      visible = true
    },
    role = "user"
  } }

I have also tried :Copilot auth signout and signin again.

[PrimaMateria]

Checking the codecompanion logs I found

Request:
{ "-sSL", "-D", "/run/user/1000//plenary_curl_be934d7b.headers", "-X", "POST", "-H", "Content-Type: application/json", "-H", "Copilot-Integration-Id: vscode-chat", "-H", "Editor-Version: Neovim/0.12.0", "-H", "Authorization: Bearer <SECRET>", "-d", "@/tmp/nvim.primamateria/c77xMv/1.json", "--retry", "3", "--retry-delay", "1", "--keepalive-time", "60", "--connect-timeout", "10", "--tcp-nodelay", "--no-buffer", "https://api.githubcopilot.com/chat/completions" }
[DEBUG] 2025-12-04 16:18:09
Output data:
bad request: invalid token: unknown stamp "prod-sdc-01"

I tried send the curl request to the https://copilot-api.finapi.ghe.com/chat/completions instead, and it failed with some different error, but at least not with that unknown stamp error - maybe I am just writing the curl command wrongly. But it seems that I need to somehow configure this URL endpoint in the adapter.

[PrimaMateria]

So far I found that the after the token is received form the github, there are endpoints defined that should be called. But the api endpoint doesn't make it to the final request when the message is sent from the chat window. I am still looking for a way how to pass it. The self.url seems to be fully ignored. Probably the config of the original public copilot plugin is in play, and we don't override it properly.

[PrimaMateria]

The main issue is that the code companion always sends request to the public endpoint, but the enterprise endpoint must be used.
This enterprise endpoint comes back when the token is retrieved.

I found the following hack, it's pretty nasty, but it works for now.
First, inside my neovim config I override the code companion's post call so it will apply the global __cc_copilot_enterprise_url when set.

  -- ENDPOINT HACK
  -- Force CC Copilot Enterprise traffic to routed GHE endpoint
  _G.__cc_copilot_enterprise_url = nil

  do
    local http = require("codecompanion.http")

    -- keep original function CC captured
    local orig_post = http.static.methods.post.default

    http.static.methods.post.default = function(opts)
      if opts
        and opts.url == "https://api.githubcopilot.com/chat/completions"
        and _G.__cc_copilot_enterprise_url
      then
        opts = vim.tbl_deep_extend("force", opts, {
          url = _G.__cc_copilot_enterprise_url,
        })
      end
      return orig_post(opts)
    end
  end
  -- END

Then I patched this plugin to set this global variable once the token is received.

---Get and authorize a GitHub Copilot Enterprise token
---@param self CodeCompanion.Adapter
---@return boolean success
---@return table|nil GitHub token on success
local function get_and_authorize_token(self)
  _oauth_token = get_token(self)
  if not _oauth_token then
    log:error(
      "Copilot Enterprise Adapter: No token found. Please ensure you've configured your Copilot Enterprise token"
    )
    return false
  end

  _github_token = authorize_token(self)
  if not _github_token or vim.tbl_isempty(_github_token) then
    log:error("Copilot Enterprise Adapter: Could not authorize your Copilot Enterprise token")
    return false
  end

  self.url = _github_token.endpoints.api:gsub("/+$", "") .. "/chat/completions"
  _G.__cc_copilot_enterprise_url = self.url
  log:debug("Copilot Enterprise API URL set to: %s", self.url)

  return true, _github_token
end

It's quite hacky solution. It would be much nicer if the code companion would dynamically resolve the adapter URL before the requests are sent, and not only once statically on the start.

[theherk]

I am struggling with a very similar:

14:33:39 │ [ERROR] 2025-12-05 14:33:31
Error: bad request: invalid token: unknown stamp "prod-weu-01"
14:33:41 │ [ERROR] 2025-12-05 14:33:31
[chat::_submit_http] Error: {
  body = 'bad request: invalid token: unknown stamp "prod-weu-01"',
  exit = 0,
  headers = { "content-security-policy: default-src 'none'; sandbox", "content-type: text/plain; charset=utf-8", "strict-transport-security: max-age=31536000", "x-content-type-options: nosniff", "date: Fri, 05 Dec 2025 13:33:31 GMT", "content-length: 56", "x-github-backend: Kubernetes", "x-github-request-id: D53F:B923:18A88F4:1CB5EBB:6932DF2B" },
  status = 400
}

Would be good to get a fix into the adapter.

[Eckii24]

I'm running into the same issue.
For me it seams like this line in the copilot adapter is the root cause:
https://github.com/olimorris/codecompanion.nvim/blob/f61dbb22439386585a514d15be713c355f830e49/lua/codecompanion/adapters/http/copilot/init.lua#L76

[mintelm]

Same issue for me, the suggested fixes of @PrimaMateria work for me. However, it seems like @dyamon is not maintaining this plugin anymore?