GuideTool IntegrationsAPI ReferenceTerms & Policies
API ReferenceText SeriesAnthropic Native ProtocolPOSTMessages
Text SeriesPOSTMessages Endpoint
API Reference

Messages

POSThttps://api.tokenbay.com/v1/messages

Anthropic-compatible Messages entry

Messages Endpoint

TokenBay currently registers an Anthropic-compatible Messages creation path.

anthropic
POST/v1/messages

Request

schema
modelstringRequired

An available Claude/Anthropic-compatible model ID shown on the live Models page or in the console.

max_tokensintegerRequired

Maximum output tokens.

messagesarray<object>Required

Anthropic Messages array. content may include text, image, tool_result, and other blocks.

messages[].rolestringRequired

user or assistant.

messages[].contentstring | array<object>Required

Text or content block array.

content[].typestringOptional

Content block type: text, image, tool_use, tool_result, thinking, and similar blocks.

content[].textstringOptional

Text content.

content[].sourceobjectOptional

Image source, usually containing type, media_type, data, or url.

content[].idstringOptional

tool_use or tool_result ID.

content[].namestringOptional

Tool name.

content[].inputobjectOptional

tool_use input.

content[].contentstring | array<object>Optional

tool_result content.

content[].cache_controlobjectOptional

Prompt caching control. Support depends on the model and upstream.

systemstring | array<object>Optional

System prompt.

system[].typestringOptional

System content block type, usually text.

system[].textstringOptional

System prompt text.

system[].cache_controlobjectOptional

System prompt cache control.

streambooleanOptional

Set true to receive Anthropic SSE events.

temperature / top_p / top_knumberOptional

Sampling controls.

stop_sequencesstring[]Optional

Sequences where generation should stop.

tools / tool_choicearray<object> | objectOptional

Anthropic tool calling configuration.

tools[].namestringOptional

Tool name.

tools[].descriptionstringOptional

Tool description.

tools[].input_schemaobjectOptional

Tool input JSON Schema.

tool_choice.typestringOptional

auto, any, tool, none, and similar modes.

tool_choice.namestringOptional

Specific tool name.

thinkingobjectOptional

Extended thinking configuration for models that support it.

thinking.typestringOptional

enabled or disabled.

thinking.budget_tokensintegerOptional

Thinking token budget.

metadataobjectOptional

Client-side tracing fields. Do not include sensitive data.

Response

schema

Non-streaming responses keep the Anthropic Messages shape. Streaming calls return Anthropic SSE events.

idstringOptional

Message ID.

typestringOptional

Usually message.

rolestringOptional

Usually assistant.

content[]array<object>Optional

Output content blocks, such as text, tool_use, or thinking.

content[].typestringOptional

Output block type, such as text, tool_use, or thinking.

content[].textstringOptional

Text output.

content[].idstringOptional

Tool call ID.

content[].namestringOptional

Tool name.

content[].inputobjectOptional

Tool input arguments.

modelstringOptional

Actual response model.

stop_reasonstringOptional

Why generation stopped.

usageobjectOptional

Input and output token usage.

usage.input_tokensintegerOptional

Input token count.

usage.output_tokensintegerOptional

Output token count.

usage.cache_creation_input_tokensintegerOptional

Cache creation token count when supported.

usage.cache_read_input_tokensintegerOptional

Cache read token count when supported.

Clients authenticate to TokenBay with Authorization: Bearer. Real upstream Anthropic credentials and related headers are handled by the gateway credential resolver and adapter layer.

Related

GeminiErrors