Aider

At the end you have aider talking to https://api.uplinkswitch.com/v1 with one API key from the console and any model id on the list after --model openai/, Claude included. Each request bills per token to the prepaid balance behind that key; no subscription, and the live table below is the rate on each model today.

You need aider installed (python -m pip install aider-install then aider-install, per its docs), an account on the console with credit on it, and one key created there. Create the key for aider alone, so its requests sit apart from any other tool on the Usage page. The base URL and the key are the whole configuration.

Aider reads an OpenAI-compatible endpoint from two environment variables, OPENAI_API_BASE and OPENAI_API_KEY, and takes the model as --model openai/<id>. The same values can live in .aider.conf.yml or .env, which aider loads from your home directory, then the root of the git repo, then the current directory; later files win (aider docs, read 2026-10-28).

The quick way is to export the two variables and run aider.

export OPENAI_API_BASE=https://api.uplinkswitch.com/v1
export OPENAI_API_KEY=sk-uplink-live-...
cd /to/your/project
aider --model openai/claude-sonnet-5-5

To keep it, put the same values in .aider.conf.yml in your home directory for every project, or in the repo root for one.

model: openai/claude-sonnet-5-5
weak-model: openai/claude-haiku-5-5
openai-api-base: https://api.uplinkswitch.com/v1
openai-api-key: sk-uplink-live-...

Or a .env file in the same places, with the names from aider's sample file: AIDER_OPENAI_API_BASE for the base URL and OPENAI_API_KEY for the key.

AIDER_OPENAI_API_BASE=https://api.uplinkswitch.com/v1
OPENAI_API_KEY=sk-uplink-live-...
AIDER_MODEL=openai/claude-sonnet-5-5

Any chat id on the Models page, written as openai/<id>. The Claude ids answer the OpenAI format as well as the Anthropic one, so openai/claude-sonnet-5-5 goes in the same flag as openai/gpt-6.1-sol or openai/deepseek-v4-pro-0813. The ids in the table below are the ones that fit an editing agent: every one takes tool calls and long context.

weak-model handles commit messages and chat history summaries and editor-model the editor tasks; both take the same openai/<id> form, and their defaults follow the main model (aider docs, read 2026-10-28).

Aider sends a repo map along with each request (aider docs, read 2026-10-28), so the input column decides what a session costs. The Claude Code pricing post works an hour, a day and a month of agent turns at these rates.

Aider passes custom headers through extra_params.extra_headers in .aider.model.settings.yml, and the special name aider/extra_params applies them to every model (aider docs, read 2026-10-28). Two headers name the app on each request's usage record:

- name: aider/extra_params
  extra_params:
    extra_headers:
      HTTP-Referer: https://uplinkswitch.ai
      X-Title: Aider

The file goes in the same three places as .aider.conf.yml, or anywhere with --model-settings-file.

Aider does not know these ids, so it prints "Unknown context window size and costs, using sane defaults" and treats the model as free with unlimited context. Requests work either way. To give it the real figures, save this as .aider.model.metadata.json next to the config file. The costs are per token at today's live rate; the limits are the developer's published ones where the catalog has them.

{
  "openai/claude-sonnet-5-5": {
    "max_input_tokens": 1000000,
    "max_tokens": 128000,
    "max_output_tokens": 128000,
    "input_cost_per_token": 0.0000012,
    "output_cost_per_token": 0.000003,
    "litellm_provider": "openai",
    "mode": "chat"
  },
  "openai/claude-opus-5-5": {
    "max_input_tokens": 1000000,
    "max_tokens": 128000,
    "max_output_tokens": 128000,
    "input_cost_per_token": 0.0000028,
    "output_cost_per_token": 0.00001,
    "litellm_provider": "openai",
    "mode": "chat"
  },
  "openai/gpt-6.1-sol": {
    "max_input_tokens": 1050000,
    "max_tokens": 128000,
    "max_output_tokens": 128000,
    "input_cost_per_token": 6e-7,
    "output_cost_per_token": 0.000003,
    "litellm_provider": "openai",
    "mode": "chat"
  },
  "openai/gemini-3.1-pro-preview": {
    "max_input_tokens": 1048576,
    "max_tokens": 65536,
    "max_output_tokens": 65536,
    "input_cost_per_token": 0.0000012,
    "output_cost_per_token": 0.0000072000000000000005,
    "litellm_provider": "openai",
    "mode": "chat"
  },
  "openai/deepseek-v4-pro-0813": {
    "max_input_tokens": 1048576,
    "max_tokens": 393216,
    "max_output_tokens": 393216,
    "input_cost_per_token": 7.92e-7,
    "output_cost_per_token": 0.0000011879999999999999,
    "litellm_provider": "openai",
    "mode": "chat"
  },
  "openai/kimi-k3": {
    "max_input_tokens": 1048576,
    "max_tokens": 1048576,
    "max_output_tokens": 1048576,
    "input_cost_per_token": 9.000000000000001e-7,
    "output_cost_per_token": 0.0000045,
    "litellm_provider": "openai",
    "mode": "chat"
  }
}

With it in place, aider's own cost line after each turn uses the live rates, and it knows the context window of the model.

From a shell, the key should list the models:

curl https://api.uplinkswitch.com/v1/models -H "Authorization: Bearer $OPENAI_API_KEY"

Then start aider and send one message. The request shows on the console Usage page within seconds, with its tokens and charge.

• 401 invalid_api_key: the key is wrong or revoked, or OPENAI_API_KEY is unset in this shell. Copy it again from the console.
• 402 insufficient_balance: the balance does not cover the request. Top up in the console.
• 402 monthly_limit_reached: the key hit the monthly spend limit you set on it. Raise it in the console, or wait for the first of the month.
• 404 model_not_found: the part after openai/ is not on the Models page. Copy the id from the page rather than typing it: a dot, a dash or a date suffix out of place is enough.
• 403 scope_denied: the key carries a model allowlist and this id is not on it.
• 429 rate_limited: the key hit its per-minute or concurrency limit. The response carries retry-after in seconds.
• "Missing these environment variables": aider could not find OPENAI_API_BASE or OPENAI_API_KEY. Export them, or put them in .env; on Windows restart the terminal after setx.
• "Unknown context window size and costs": a warning, not an error. The metadata file above removes it.

Claude Code takes two environment variables; OpenCode and OpenClaw read a provider block; Cline and Kilo Code take the base URL and key in a settings panel. One balance serves all of them, and a key per tool keeps the spend apart in the console.

Get a key in the console: prepaid credit, no subscription, 30-day price lock.