omp

At the end you have omp, the terminal coding agent from the oh-my-pi project, on one provider block at https://api.uplinkswitch.com/v1, with one API key from the console and any model id on the list, Claude included. Each request bills per token to the prepaid balance behind that key, no subscription, and the live table below is what one agent step costs on each model today.

You need omp installed (curl -fsSL https://omp.sh/install | sh on macOS and Linux), an account on the console with credit on it, and one key created there. Create the key for omp alone, so its requests sit apart from any other tool on the Usage page. Nothing is installed on the UplinkSwitch side: the base URL and the key are the whole configuration.

Custom providers live in ~/.omp/agent/models.yml (models.yaml is read only when the .yml file is absent). Its root holds providers and nothing else.

The default model is a separate setting: modelRoles.default in ~/.omp/agent/config.yml for every project, or <repo>/.omp/config.yml for one. omp reads the project file from the directory you launch it in and does not search parent directories. Read 2026-10-28 from omp.sh/docs/custom-models and docs/settings.

api: openai-completions sends /v1/chat/completions to baseUrl, which is the format every chat model here answers. apiKey is the name of an environment variable, not the key itself. The two headers let the gateway attribute the traffic to omp on the Usage page. The limits under each model are the developer's published figures from the live catalog, so omp trims the conversation before the model refuses it.

# ~/.omp/agent/models.yml
providers:
  uplinkswitch:
    baseUrl: https://api.uplinkswitch.com/v1
    api: openai-completions
    apiKey: UPLINKSWITCH_API_KEY
    headers:
      HTTP-Referer: https://uplinkswitch.ai
      X-Title: omp
    models:
      - id: claude-sonnet-5-5
        name: Claude Sonnet 5.5
        input: [text, image]
        contextWindow: 1000000
        maxTokens: 128000
      - id: claude-opus-5-5
        name: Claude Opus 5.5
        input: [text, image]
        contextWindow: 1000000
        maxTokens: 128000
      - id: gpt-6.1-sol
        name: GPT-6.1 Sol
        input: [text, image]
        contextWindow: 1050000
        maxTokens: 128000
      - id: gemini-3.1-pro-preview
        name: Gemini 3.1 Pro Preview
        input: [text, image]
        contextWindow: 1048576
        maxTokens: 65536
      - id: deepseek-v4-pro-0813
        name: DeepSeek V4 Pro 0813
        contextWindow: 1048576
        maxTokens: 393216
      - id: kimi-k3
        name: Kimi K3
        input: [text, image]
        contextWindow: 1048576
        maxTokens: 1048576

Every id under models shows up in /model as uplinkswitch/<id>. Any id on the Models page can be added the same way. A model entry can also carry reasoning: true and a thinking block for effort controls, and a cost block with input, output, cacheRead and cacheWrite per million tokens for omp's own cost line; omp needs all four, and the console bills the live price either way.

Export the variable in the shell that launches omp, or put it in a .env file. omp loads .env from the current directory, ~/.omp/agent, ~/.omp and your home directory, in that order, with an exported value winning over all of them.

export UPLINKSWITCH_API_KEY="sk-uplink-live-..."
omp

# ~/.omp/agent/.env
UPLINKSWITCH_API_KEY=sk-uplink-live-...

A value starting with ! runs a shell command and uses its output as the key, for a secret manager. --api-key on the command line overrides apiKey for one run and is not saved.

Tool calling works on every chat model here, and omp's agent needs it for every file read, edit and command, so no id in the block is a dead end. Input price decides the session cost, because omp sends the conversation and every tool result back on each step; pick by the input column below. Set the default once:

# ~/.omp/agent/config.yml (or <repo>/.omp/config.yml)
modelRoles:
  default: uplinkswitch/claude-sonnet-5-5

Or start omp, run /model and assign a model to Default. /switch (Alt+P) changes only the current session; --model on the command line wins over everything. --slow, --smol and --plan override the thorough, lightweight and planning roles with any uplinkswitch/<id>. The uncensored coding builds go in the same block, billed per token.

One agent step at 40,000 input tokens and 1,500 output tokens, which is a task that has read a few files and run a command or two. Nothing is counted at the cached rate.

Claude Sonnet 5.5 is $0.0525 for that step here, and a ten-step task is $0.525. Cached input bills at the cache-read rate in the table by itself; there is no switch to set. The Claude Code pricing post works a day and a month of agent steps at these rates.

omp models uplinkswitch lists the ids from the block; a bad field prints models.yml validation failed with its name. The second line sends one request in print mode:

omp models uplinkswitch
omp -p --model uplinkswitch/claude-sonnet-5-5 "Reply with only OK"

From a shell, the same key should list the models:

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

The request shows on the console Usage page within seconds, with its tokens and its charge, recorded with omp as the app.

• 401 invalid_api_key: the key is wrong or revoked, or the variable named in apiKey is not set in the shell that launched omp. A name omp cannot find is sent as the literal key. Export it, or check the .env path, and copy the key 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 id under models is not on the Models page at this base URL. Copy it 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. Pick an allowed id or use a key without the limit.
• 429 rate_limited: the key hit its requests-per-minute or concurrency limit. The response carries retry-after in seconds. Wait that long and resend.
• The provider is missing from /model: the file is models.yml, not config.yml, and baseUrl ends in /v1. Run omp models uplinkswitch for the failing field.

• Only the ids you list are offered. discovery with type: openai-models-list can read GET /v1/models instead, which brings in every chat id at once; omp models refresh uplinkswitch re-reads it.
• A model omitted from a field defaults to text-only input, a 128,000 context window and 16,384 output tokens, which is why the block above fills them from the catalog.
• --models restricts what Ctrl+P cycles through, by the same provider/id selectors.
• The same key works in OpenCode, Claude Code and Cline on the same balance. A key per tool keeps the spend apart on the Usage page.

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