Which API should you use with Rakazo?

For Rakazo, the self-hosted alternative to Grok Bot, Meta Muse and OpenAI Dots, the best API for bots that run routines all day is a cost-optimized OpenAI-compatible endpoint: set RAKAZO_OPENAI_COMPAT_ALLOW_PUBLIC=1 in your deployment, then in Settings → Models choose OpenAI-compatible with the server URL https://api.stdcmpt.com/v1 and the Model id standardcompute, and Standard Compute's router sends each step to the cheapest model that can do it (Claude Fable 5.1 or GPT-6 Astra only when it counts), on a plan from $19 a month with a $20 compute budget. OpenRouter and direct OpenAI or Anthropic keys bill per token, and a ChatGPT plan can sign in for OpenAI's models. Rakazo itself is free and Apache-2.0.

Light or occasional use

If you already pay for ChatGPT and your bots fit inside its Codex allowance, Rakazo's ChatGPT sign-in adds no new bill. For one bot with a few routines a day, an OpenRouter key on an efficient model will likely cost the least.

Daily use, cost-optimized

When several bots run routines, browse and work on their computers every day, a Standard Compute endpoint sends each step to the cheapest model that can do it, and for every model we vet providers for price, uptime and speed, with automatic fallback. Opt-in Europe or US Compliance rules apply per key, team or company, and we don't train on your prompts or store them. The bill stays at the monthly plan price: requests stop with HTTP 402 when the compute budget is used, so a looping bot cannot bill you on top.

API and subscription options for Rakazo

Prices and allowances can change. Rakazo's OpenAI-compatible connection, the public-endpoint flag and the Supports images and Supports thinking switches come from its self-hosting guide and model settings source, read on October 15, 2026. Rakazo: self-hosting guide · Rakazo repo · Rakazo: Grok Bot alternative. Spot an error? Tell us.

Connect Rakazo to Standard Compute

# .env of your Rakazo deployment (then recreate the stack)
RAKAZO_OPENAI_COMPAT_ALLOW_PUBLIC=1

# Settings → Models → OpenAI-compatible
OpenAI-compatible server URL  https://api.stdcmpt.com/v1
Model id                      standardcompute
API key                       sc_live_…   (your Standard Compute key)
Advanced                      Supports images: on

Without the flag, Rakazo refuses public model endpoints; a public endpoint must also use HTTPS. Pick Models from server instead of typing to choose an exact ID such as anthropic/claude-sonnet-5.5. Each bot can use a different connected model.

Rakazo questions

Why does Rakazo say public model endpoints are blocked?

Rakazo only lets its OpenAI-compatible connection reach private, loopback or host.docker.internal addresses by default. For a hosted API such as https://api.stdcmpt.com/v1, the deployment owner sets RAKAZO_OPENAI_COMPAT_ALLOW_PUBLIC=1 in .env and recreates the stack; the endpoint must use HTTPS. Prompts, attachments and tool results then leave Rakazo through that URL, so only configure one you trust.

Can Rakazo run on the same models as Grok Bot or Meta Muse?

No. Grok Bot and Muse run on their makers' own models and have no model picker or API key setting. Rakazo is the open-source way to get persistent bots with their own computers on a model you choose: any OpenAI-compatible endpoint, OpenAI, Anthropic, Google, OpenRouter, or a ChatGPT sign-in. With Standard Compute's model standardcompute, the router picks the cheapest model that can do each step.

Do Rakazo's computer tools work with a custom endpoint?

Yes, when the model accepts images. Under Advanced, turn on Supports images when you connect, so attachments and the screenshot computer tools stay available; existing connections default to off. Turn on Supports thinking only if the endpoint accepts the standard reasoning_effort parameter.

What's the best API for an open-source Grok Bot alternative that runs bots 24/7 without a surprise bill?

One with a hard monthly budget. On Standard Compute, Rakazo's bots can use at most your plan's compute budget (from $19 a month with a $20 compute budget): at 100% new requests stop with HTTP 402 until renewal, with nothing billed on top, and the model standardcompute sends routine steps to cheap models so the budget lasts. OpenRouter and direct keys bill per token, so a bot stuck in a loop costs whatever it burns before its credits or key limit run out.