Any LLMAPI
DEVELOPERS

Any LLM API documentation for developers

Protocols, request examples and integration limits.

Loading API settings…

Clients and model selection

The connector reads /v1/models, keeps models with tools and a compatible protocol, and launches your installed client. It does not save the key to disk. Run the command again to update the catalog.

  • Hermes: /model · generated catalog in a separate profile
  • OpenCode 1.14.50: /models · Linux flow checked; OpenCode 2.x is not verified yet
  • Codex CLI 0.163.0: /model · Linux CLI with Responses catalog checked; Codex IDE/desktop not checked
  • Claude Code CLI 2.1.306: /model · Linux flow checked; the full model list requires a configured modelPicker

The connector prepares a separate profile and loads the catalog from /v1/models. For another client or version, follow its documentation and verify the required features. A model in the catalog does not guarantee every feature or route.

Hermes 09e2ffc and the listed CLI versions were also checked on Linux. Windows, macOS, OpenCode 2.x and Codex IDE/desktop require separate checks.

Billing and free access

Free models need no subscription or balance, within shared capacity and quotas. PRO uses its weekly money budget; MAX uses its monthly budget. Paid requests use the subscription budget, then the wallet. Your bank card is not charged automatically.

The estimated maximum cost is reserved before a request. Unused funds are released after settlement. Unknown outcomes keep their reservation until reconciled. Partial content and reasoning are billable. Owner-defined key limits and abuse protections apply to every key.

Current quotas and trial remainder appear in the API account and depend on server configuration and the account. Open API account

Supported protocols

  • POST /v1/chat/completions — Chat Completions, JSON/SSE.
  • POST /v1/responses — native Responses API.
  • POST /v1/messages — Anthropic Messages.

Check the protocols listed for each model in the catalog. Tools, streaming, reasoning and request parameters depend on the model and route. OpenAPI

Fetch the catalog and use the exact ID of a selected model from the response:

curl https://api.any-llm.online/v1/models \
  -H 'Authorization: Bearer YOUR_API_KEY'

curl https://api.any-llm.online/v1/chat/completions \
  -H 'Authorization: Bearer YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{"model":"EXACT_MODEL_ID_FROM_V1_MODELS","messages":[{"role":"user","content":"Hello"}],"max_tokens":100,"store":false}'

Replace placeholders locally. Do not publish a real key or automatically retry a generation request with an unknown outcome.

Chat Completions and Messages do not store content by default; store:true enables history for up to 30 days. Responses stores history by default; store:false disables it. Closing a stream does not cancel work: use explicit cancellation.

401: key; 402: insufficient funds; 429: quota or capacity; 503: temporary unavailability. Do not automatically retry requests with unknown outcomes. Keep X-Request-ID for support.

Available API models