FAQ

Q: When should I not use Boundless?

Do not reach for it when you need provider-native Responses features such as OpenAI-hosted tools or remote conversation storage. /v1/responses serves the Codex-compatible subset, not those hosted features. Do not use it for embeddings, legacy completions (/v1/completions), fine-tuning, or image or audio generation — none of these are served. Do not use it if you need a specific proprietary model: only open-weight models are served, and Anthropic's claude-* and OpenAI's model names are not available here under any alias. Sign-up and key creation are self-serve, but the service is invite-only and a human approves each new account while the first accounts are onboarded.

Q: Can I try the API without an account?

Yes. POST /api/agent-key with no body and no credential, or use the create_agent_key MCP tool. The key is a smoke-test credential: $0.01 of credit, expires in 24 hours, limited to 6 requests and 30,000 tokens per minute. It cannot be topped up or extended, and it is not for a real workload. GET on the same URL describes the offer without minting. Keys for real work are created in the console once an account is active.

Q: Which models does Boundless serve?

Only open-weight models are served. Anthropic's claude-* and OpenAI's model names are not available here under any alias. Each model's API identifier, context window, capabilities, and live rates are on its catalog page. Prices are deliberately not restated here: a price copied into a second place is a price that can be wrong.

Q: What does Boundless not serve?

Embeddings (/v1/embeddings), legacy completions (/v1/completions), fine-tuning, image generation, and audio generation are not served. /v1/responses is the Codex-compatible subset only — not OpenAI-hosted tools or remote conversation storage. Batch inference with 1h and 24h completion windows launches in the coming weeks; the discounted rates are on the pricing page.

When to use Boundless, when not to, the wire formats, prepaid credits, and what is not served.

Short answers to the questions the docs already imply. The quickstart is the path to a first token; llms.txt is the same fit guidance for an agent.

When should I use Boundless?#

Reach for Boundless when you already call an OpenAI or Anthropic chat API and want the same code to run open-weight models — the wire formats are the ones your SDK already sends, so the change is a base URL and a key.

Use it when you need:

the exact cost of each request, not a monthly total — every response carries token counts that reproduce the charge against the rate card
a hard spending ceiling — credits are prepaid, requests stop at zero, and a key can carry its own budget
a coding agent or harness pointed at cheaper models without changing the harness — Codex CLI, Claude Code, omp, Hermes, OpenCode, and others
long context, or tool use and reasoning on open weights

When should I not use Boundless?#

Do not reach for it when:

you need provider-native Responses features such as OpenAI-hosted tools or remote conversation storage. /v1/responses serves the Codex-compatible subset, not those hosted features
you need embeddings, legacy completions (/v1/completions), fine-tuning, or image or audio generation. None of these are served
you need a specific proprietary model. Only open-weight models are served; Anthropic's claude-* and OpenAI's model names are not available here under any alias
you need instant self-service activation today. Sign-up and key creation are self-serve, but the service is invite-only and a human approves each new account while the first accounts are onboarded

A capped agent key is mintable instantly for smoke tests; a human account is only needed for real spend.

Which wire formats does the API speak?#

One service speaks three wire formats — OpenAI chat completions, the Codex-compatible Responses subset, and the Anthropic Messages API — over the same models on the same rate card.

An OpenAI SDK or Codex custom provider takes the gateway /v1 base.
Anthropic SDKs take the gateway root — they append /v1/messages themselves.
Codex uses wire_api = "responses".

Authenticate with Authorization: Bearer, or with x-api-key if that is what your client already sends. Name a model from the catalog. See API Compatibility.

How do prepaid credits work?#

Access is sold on prepaid credits. Requests debit the shared Team balance exactly. Every charge can be recomputed from the rate card and each response's usage object. A single API key can carry its own budget below the team's. Only Team admins can buy credits or change the card on file.

What happens when credits reach zero?#

Requests stop. There is no overage and no end-of-month invoice. Exhausted credits return 429 budget_exceeded and are not charged past the balance. Waiting does not recover spend — top up. The console warns as the balance runs low.

Can I try the API without an account?#

Yes. POST /api/agent-key with no body and no credential, or use the create_agent_key MCP tool. The key is a smoke-test credential: it carries a small cap, expires, and cannot be topped up or extended. It is not for a real workload. GET on the same URL describes the offer without minting.

Keys for real work are created in the console once an account is active.

Is the service invite-only?#

Yes, while the first accounts are onboarded. Sign-up and key creation are self-serve, but a human approves each new account. The model lineup, prices, and base URL may change during that period.

Which models does Boundless serve?#

Only open-weight models are served. Anthropic's claude-* and OpenAI's model names are not available here under any alias.

Each model's API identifier, context window, capabilities, and live rates are on Models. Prices are deliberately not restated here: a price copied into a second place is a price that can be wrong.

How long is the context window?#

Each model publishes its own window on its catalog page. Several served models take 262,144 tokens; others advertise longer. Prompt and completion share that window. Check the live model page rather than a copied number.

What does Boundless not serve?#

API Compatibility keeps the surface small:

embeddings (/v1/embeddings)
legacy completions (/v1/completions)
fine-tuning
image generation and audio generation
Responses features outside the Codex-compatible subset — not OpenAI-hosted tools or remote conversation storage

Batch inference with 1h and 24h completion windows launches in the coming weeks; the discounted rates are on Pricing.

Can I point a coding agent or harness at Boundless?#

Yes. Codex CLI, Claude Code, omp, Hermes, OpenCode, and other clients that let you set a base URL and an API key work here. The client's wire format decides which base URL you give it. See Coding Agents & Harnesses.

FAQ