Create message (Anthropic)

/v1/messages

Examples verified against code, not live (TI-78). Send the key as x-api-key (what the Anthropic SDKs send) or Authorization: Bearer; anthropic-version is accepted with any value. Errors wear Anthropic's envelope {type:'error', error:{type, message}, request_id}, on /v1/messages and /api/v1/messages alike.

Streaming (stream:true): message_start, then per content block content_block_start, content_block_delta (text_delta or input_json_delta) and content_block_stop, then message_delta (stop reason and usage) and message_stop. ping is the keepalive; a failure after the 200 ends the stream with an error event.

The anthropic-beta header is not forwarded upstream, so beta-gated features do not switch on. Responses carry text and tool_use blocks only.

Authorizations

Authorizationstringheaderrequired

Bearer token: Authorization: Bearer …. ozk_ key in the x-api-key header, as the Anthropic SDKs send it — POST /v1/messages only (Authorization: Bearer also works there and wins when both are sent). anthropic-version is accepted with any value.

Authorizationstringheaderrequired

Bearer token: Authorization: Bearer ozk_…. ozk_ platform API key (Authorization: Bearer ozk_…)

Authorizationstringheaderrequired

Bearer token: Authorization: Bearer sk-ocean-…. Legacy sk-ocean- inference key: still accepted on the money plane until the sk-ocean- sunset; new integrations use the ozk_ key

Body

application/json

Anthropic Messages request (POST /v1/messages), any model slug. On an anthropic/* model the body is forwarded as sent, with only model (the upstream id) and max_tokens (the capped value) set by the gateway, so cache_control, thinking, documents and server tools reach Anthropic unchanged. On any other model it is translated to the chat-completions shape; blocks with no neutral form (thinking, server tools) are not sent. Gateway directives (models, fallbacks, sort, max_cost and OpenRouter's provider, route, transforms, plugins, usage, session_id) are consumed, never forwarded.

modelstring

A model slug ("provider/model") or "@preset/<slug>". Required unless models[] is set.

Example: "anthropic/claude-haiku-4-5"

messagesobject[]required

Turns, role user or assistant. content is a string or an array of blocks: text, image (source base64 or url), tool_use, tool_result, thinking, document, … tool_result blocks become tool messages on a translated model.

Example: [{"role":"user","content":"Say hi in three words"}]

Show child attributes

roleenum<string>required

Available options: user, assistant

contentstring | AnthropicContentBlock[]required

Option 1

typestringrequired

Example: "text"

textstring

sourceobject

Show child attributes

typeenum<string>

Available options: base64, url

media_typestring

datastring

urlstring

idstring

namestring

inputobject

tool_use_idstring

contentstring | object[]

tool_result only: a string or an array of text blocks

Option 1

Option 2

is_errorboolean

cache_controlobject

systemstring | AnthropicContentBlock[]

A string or an array of text blocks; a system message on a translated model.

Option 1

typestringrequired

Example: "text"

textstring

sourceobject

Show child attributes

typeenum<string>

Available options: base64, url

media_typestring

datastring

urlstring

idstring

namestring

inputobject

tool_use_idstring

contentstring | object[]

tool_result only: a string or an array of text blocks

Option 1

Option 2

is_errorboolean

cache_controlobject

max_tokensinteger

Example: 64

stop_sequencesstring[]

stop on a translated model

streamboolean

Default: false

temperaturenumber

top_pnumber

top_kinteger

toolsobject[]

Client tools (name, description, input_schema → function.parameters on a translated model). A server tool (a type other than custom, no input_schema) only an anthropic/* model can run.

Show child attributes

namestringrequired

descriptionstring

input_schemaobject

typestring

tool_choiceobject

On a translated model: auto → auto, any → required, none → none, tool → that function; disable_parallel_tool_use → parallel_tool_calls:false.

Show child attributes

typeenum<string>required

Available options: auto, any, tool, none

namestring

disable_parallel_tool_useboolean

metadataobject

Show child attributes

user_idstring

user on a translated model

thinkingobject

Forwarded to anthropic/*. On a translated model, type enabled with budget_tokens becomes reasoning_effort by its share of max_tokens: ≤35% low, ≤65% medium, else high.

Show child attributes

typestring

budget_tokensinteger

service_tierstring

standard_only is dropped and auto is forwarded (both the default tier); a value other than auto or default answers 400 unsupported_service_tier (the rate card prices the default tier only).

modelsstring[]

Ocean: ordered fallback candidates, as on /v1/chat/completions.

fallbacksstring[]

OpenRouter's Anthropic-skin fallback list: tried after model, at most 3 (more → 400). Ignored when models[] is set.

Maximum array length: 3

sortenum<string>

Available options: price, latency, throughput

max_costnumber

Ocean: the request’s own cost cap in USD.

Response

An Anthropic message (model = the slug that actually served it). x-request-id and the x-ocean-* headers are kept. stream:true returns text/event-stream: message_start, then per block content_block_start / content_block_delta (text_delta, input_json_delta) / content_block_stop, then message_delta (stop_reason and usage) and message_stop; ping is the keepalive, and a failure after the 200 ends the stream with an error event.

application/json
text/event-stream

Response headers

x-request-idstringfrom code

Server-minted request id; also request_id in an error body.

idstringrequired

Example: "msg_0f3c…"

typeenum<string>required

Available options: message

roleenum<string>required

Available options: assistant

modelstringrequired

contentobject[]required

Show child attributes

typeenum<string>required

Available options: text, tool_use

textstring

idstring

namestring

inputobject

stop_reasonenum<string>required

An anthropic/* upstream's own stop reason; otherwise from finish_reason (length → max_tokens, tool_calls → tool_use, content_filter → refusal, else end_turn).

Available options: end_turn, max_tokens, stop_sequence, tool_use, pause_turn, refusal

stop_sequenceany | nullrequired

usageobjectrequired

Show child attributes

input_tokensinteger

Excludes cache reads and cache writes

output_tokensinteger

cache_creation_input_tokensinteger

cache_read_input_tokensinteger

invalid_request_error — invalid_json (body is not valid JSON), messages required, a role other than user/assistant, more than 3 fallbacks, or unsupported_service_tier

application/json

typeenum<string>required

Available options: error

errorobjectrequired

Show child attributes

typestringrequired

Example: "authentication_error"

messagestringrequired

request_idstring

Equals the x-request-id header

authentication_error — invalid_api_key: no x-api-key or Authorization: Bearer header, or not a valid key

application/json

typeenum<string>required

Available options: error

errorobjectrequired

Show child attributes

typestringrequired

Example: "authentication_error"

messagestringrequired

request_idstring

Equals the x-request-id header

billing_error — insufficient_balance (the wallet does not cover the request) or byok_required

application/json

typeenum<string>required

Available options: error

errorobjectrequired

Show child attributes

typestringrequired

Example: "authentication_error"

messagestringrequired

request_idstring

Equals the x-request-id header

permission_error — model_not_allowed (allow-list), content_blocked (guardrail) or an inactive account

application/json

typeenum<string>required

Available options: error

errorobjectrequired

Show child attributes

typestringrequired

Example: "authentication_error"

messagestringrequired

request_idstring

Equals the x-request-id header

not_found_error — model_not_found or preset_not_found

application/json

typeenum<string>required

Available options: error

errorobjectrequired

Show child attributes

typestringrequired

Example: "authentication_error"

messagestringrequired

request_idstring

Equals the x-request-id header

method_not_allowed — the path serves one method: the Allow header names it, plus OPTIONS

application/json

Response headers

Allowstring

The method this path serves, plus OPTIONS

errorobjectrequired

Show child attributes

codestringrequired

messagestringrequired

connectobject

On a 424 connection_required / credential_required: where the missing connection is made.

Show child attributes

providerstring

methodenum<string>

Available options: oauth, api_key, wallet

linkstring

Example: "POST /v1/connections/stripe/link"

rate_limit_error — rate_limit_exceeded, spend_cap_exceeded or insufficient_quota

application/json

typeenum<string>required

Available options: error

errorobjectrequired

Show child attributes

typestringrequired

Example: "authentication_error"

messagestringrequired

request_idstring

Equals the x-request-id header

api_error — every routing candidate failed

application/json

typeenum<string>required

Available options: error

errorobjectrequired

Show child attributes

typestringrequired

Example: "authentication_error"

messagestringrequired

request_idstring

Equals the x-request-id header

money_plane_unavailable — INFERENCE_GATEWAY_SERVICE not bound on this tier

application/json

errorobjectrequired

Show child attributes

codestringrequired

messagestringrequired

connectobject

On a 424 connection_required / credential_required: where the missing connection is made.

Show child attributes

providerstring

methodenum<string>

Available options: oauth, api_key, wallet

linkstring

Example: "POST /v1/connections/stripe/link"

import os import anthropic client = anthropic.Anthropic(base_url='https://api.plungeai.com', api_key=os.environ['PLUNGE_API_KEY']) message = client.messages.create( model='anthropic/claude-haiku-4-5', max_tokens=64, messages=[{'role': 'user', 'content': 'hi'}], ) print(message.content[0].text)

import os import anthropic client = anthropic.Anthropic(base_url='https://api.plungeai.com', api_key=os.environ['PLUNGE_API_KEY']) with client.messages.stream( model='openai/gpt-4o-mini', max_tokens=64, messages=[{'role': 'user', 'content': 'hi'}], ) as stream: for text in stream.text_stream: print(text, end='', flush=True)

import Anthropic from '@anthropic-ai/sdk' const client = new Anthropic({ baseURL: 'https://api.plungeai.com', apiKey: process.env.PLUNGE_API_KEY }) const message = await client.messages.create({ model: 'anthropic/claude-haiku-4-5', max_tokens: 64, messages: [{ role: 'user', content: 'hi' }], }) console.log(message.content[0].type === 'text' ? message.content[0].text : message.content)

import Anthropic from '@anthropic-ai/sdk' const client = new Anthropic({ baseURL: 'https://api.plungeai.com', apiKey: process.env.PLUNGE_API_KEY }) const stream = client.messages.stream({ model: 'openai/gpt-4o-mini', max_tokens: 64, messages: [{ role: 'user', content: 'hi' }], }) stream.on('text', (text) => process.stdout.write(text)) await stream.finalMessage()

import java.net.URI; import java.net.http.HttpClient; import java.net.http.HttpRequest; import java.net.http.HttpResponse; public class Main { public static void main(String[] args) throws Exception { String json = "{\"model\":\"anthropic/claude-haiku-4-5\",\"max_tokens\":64,\"messages\":[{\"role\":\"user\",\"content\":\"hi\"}]}"; HttpRequest req = HttpRequest.newBuilder(URI.create("https://api.plungeai.com/v1/messages")) .header("x-api-key", System.getenv("PLUNGE_API_KEY")) .header("anthropic-version", "2023-06-01") .header("Content-Type", "application/json") .POST(HttpRequest.BodyPublishers.ofString(json)) .build(); HttpResponse<String> res = HttpClient.newHttpClient().send(req, HttpResponse.BodyHandlers.ofString()); System.out.println(res.statusCode() + " " + res.body()); } }

import java.net.URI; import java.net.http.HttpClient; import java.net.http.HttpRequest; import java.net.http.HttpResponse; public class Main { public static void main(String[] args) throws Exception { HttpClient client = HttpClient.newHttpClient(); HttpRequest request = HttpRequest.newBuilder() .uri(URI.create("https://api.plungeai.com/v1/messages")) .header("Authorization", "Bearer " + System.getenv("PLUNGE_API_KEY")) .header("Content-Type", "application/json") .POST(HttpRequest.BodyPublishers.ofString("{\"messages\":[{\"role\":\"user\",\"content\":\"hi\"}],\"stream\":true}")) .build(); HttpResponse<java.util.stream.Stream<String>> response = client.send(request, HttpResponse.BodyHandlers.ofLines()); if (response.statusCode() >= 400) { // Models plane: the machine code is error.code in the JSON body. System.out.println(response.statusCode() + " " + String.join("\n", response.body().toList())); return; } response.body().takeWhile(line -> !line.equals("data: [DONE]")).forEach(System.out::println); } }

curl -s https://api.plungeai.com/v1/messages \ -H "x-api-key: $PLUNGE_API_KEY" \ -H 'anthropic-version: 2023-06-01' \ -H 'Content-Type: application/json' \ -d '{"model":"anthropic/claude-haiku-4-5","max_tokens":64,"messages":[{"role":"user","content":"hi"}]}'

curl -sN https://api.plungeai.com/v1/messages \ -H "x-api-key: $PLUNGE_API_KEY" \ -H 'anthropic-version: 2023-06-01' \ -H 'Content-Type: application/json' \ -d '{"model":"openai/gpt-4o-mini","max_tokens":64,"stream":true,"messages":[{"role":"user","content":"hi"}]}'

{ "id": "msg_0f3c…", "type": "message", "role": "assistant", "model": "<string>", "content": [ { "type": "text", "text": "<string>", "id": "<string>", "name": "<string>", "input": {} } ], "stop_reason": "end_turn", "stop_sequence": null, "usage": { "input_tokens": 0, "output_tokens": 0, "cache_creation_input_tokens": 0, "cache_read_input_tokens": 0 } }

Create message (Anthropic)

Authorizations

Body

Response

Response headers

Response headers

Planned: TI-33

Planned: TI-34