> ## Documentation Index > Fetch the complete documentation index at: https://docs.usehasp.com/llms.txt > Use this file to discover all available pages before exploring further. # Handle an Anthropic Messages-compatible request, streaming or returning the assembled response ## OpenAPI ````yaml /openapi/v1.json post /messages openapi: 3.1.0 info: title: HASP AI API version: '2026-07-12' description: >- The HASP Public AI API is the regulated-AI substrate for healthcare and other regulated industries — identity, policy, audit, compliance, and PHI handling for AI inference, exposed via two surfaces: - **Native (`/v1/ai/*`)**: HASP-native chat with full event taxonomy, run lifecycle, and PHI metadata. - **Anthropic-compat (`/v1/messages`)**: Drop-in replacement for `@anthropic-ai/sdk` — change only `baseURL`. All Gateway compliance checks (BAA, credits, PHI policy) apply. All requests require an API key (`Authorization: Bearer hasp_api_live_`). See [Authentication](https://docs.usehasp.com/ai-api/authentication) for key management. servers: - url: https://api.usehasp.com/v1 description: Production security: - http: [] paths: /messages: post: tags: - Messages summary: >- Handle an Anthropic Messages-compatible request, streaming or returning the assembled response operationId: v1.messages requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/MessagesRequest' responses: '200': description: '' content: application/json: schema: anyOf: - type: object properties: id: type: string type: type: string const: message role: type: string const: assistant content: type: array items: anyOf: - type: object properties: type: type: string const: text text: type: string required: - type - text - type: object properties: type: type: string const: tool_use id: anyOf: - type: string - type: string enum: - '' name: anyOf: - type: string - type: string enum: - '' input: anyOf: - type: string - type: array items: type: string minItems: 0 maxItems: 0 additionalItems: false required: - type - id - name - input model: {} stop_reason: type: string stop_sequence: type: 'null' usage: type: object properties: input_tokens: type: integer output_tokens: type: integer cache_creation_input_tokens: type: integer cache_read_input_tokens: type: integer required: - input_tokens - output_tokens - cache_creation_input_tokens - cache_read_input_tokens meta: type: object properties: request_id: type: - string - 'null' required: - request_id required: - id - type - role - content - model - stop_reason - stop_sequence - usage - meta - type: string text/event-stream: schema: type: string headers: Transfer-Encoding: required: true schema: type: string enum: - chunked '400': description: '' content: application/json: schema: type: object properties: type: type: string const: error error: type: object properties: type: type: string const: invalid_request_error message: type: string hasp_code: type: string const: UNSUPPORTED_PARAMETER hasp_details: type: object properties: parameter: type: string const: store required: - parameter param: type: string const: store required: - type - message - hasp_code - hasp_details - param required: - type - error '401': description: >- Anthropic-shaped error — bearer token missing, malformed, or revoked. content: application/json: schema: type: string examples: - type: error error: type: authentication_error message: Bearer token is missing, malformed, or revoked. hasp_code: INVALID_API_KEY hasp_request_id: 01JQREQ7XZQK5N6PZ1VVXHYB8T '403': $ref: '#/components/responses/AuthorizationException' '422': $ref: '#/components/responses/ValidationException' '429': description: Anthropic-shaped error — rate limit exceeded. content: application/json: schema: type: string examples: - type: error error: type: rate_limit_error message: Rate limit exceeded. hasp_code: RATE_LIMITED hasp_request_id: 01JQREQ7XZQK5N6PZ1VVXHYB8T components: schemas: MessagesRequest: type: object description: >- Validates POST /v1/messages (Anthropic-compat) per PRD §05-public-api. Accepts Anthropic's exact wire format so existing @anthropic-ai/sdk integrations (including coding agents such as Claude Code, per ADR-YGE00M) need only change the baseURL: structured content blocks (text / tool_use / tool_result / image), tool definitions, and `cache_control` markers on system blocks, content blocks, and tool definitions. HASP-native flags (e.g. 'store') are rejected with 400 UNSUPPORTED_PARAMETER per PRD §05: "rejecting rather than silently accepting keeps the contract honest." properties: model: type: string maxLength: 100 max_tokens: type: - integer - 'null' minimum: 1 maximum: 32000 tool_choice: type: - string - 'null' description: >- Shape-validated here; normalized by AnthropicMessagesRequestBuilder::buildToolChoice() and enforced by AiGatewayService::assertParametersSupported() (HASP-509). stream: type: - boolean - 'null' temperature: type: - number - 'null' minimum: 0 maximum: 1 top_p: type: - number - 'null' minimum: 0 maximum: 1 top_k: type: - integer - 'null' minimum: 1 provider_extensions: type: - array - 'null' description: |- Escape hatch for Anthropic-specific wire keys the canonical fields above don't (yet) model — see ADR-ZP99PX. Merged into the raw wire payload by DirectProviderDriver; rejected with UNSUPPORTED_PARAMETER on drivers that can't honor it. items: type: string stop_sequences: type: - array - 'null' items: type: string metadata: type: object properties: user_id: type: - string - 'null' maxLength: 255 on_behalf_of: type: object description: |- Asserted end-user identity — a HASP extension on the Anthropic wire shape (Anthropic SDK users send it via `extra_body`). Records which user of your application took this action on the audit chain, as an assertion made by your integration — HASP does not verify it, it is never forwarded to the model provider, and it is never shown to the model. Distinct from `metadata.user_id`, which is provider metadata. Supported for API-key credentials only. properties: id: type: string maxLength: 256 display: type: - string - 'null' maxLength: 256 messages: type: array items: type: object properties: role: type: string enum: - user - assistant content: type: string description: >- Content is either a plain string or an array of content blocks (text / tool_use / tool_result / image / document) — Anthropic's wire format allows both; shape-level validation of individual blocks is intentionally loose here (mirrors ChatCompletionsRequest's precedent of validating for presence, not deep block schemas). cache_control lives on individual content blocks (Anthropic's wire convention), not on the message itself — captured verbatim as part of this opaque blob when content is a block array. Restricted to string|array (not int/bool/etc.) so a malformed request 422s instead of 500ing when Message::$content type-checks it. required: - role - content minItems: 1 system: type: array description: |- System is either a plain string or an array of blocks carrying per-block cache_control (Anthropic's system-prompt caching shape). items: type: object properties: type: type: - string - 'null' enum: - text text: type: - string - 'null' cache_control: type: - object - 'null' properties: type: type: - string - 'null' enum: - ephemeral tools: type: - array - 'null' items: type: object properties: name: type: string description: type: - string - 'null' input_schema: type: - object - 'null' properties: type: type: - string - 'null' description: >- Nested key so the generated OpenAPI spec renders input_schema as a JSON Schema object rather than a bare array. cache_control: type: - object - 'null' properties: type: type: - string - 'null' enum: - ephemeral required: - model - messages title: MessagesRequest responses: AuthorizationException: description: Authorization error content: application/json: schema: type: object properties: message: type: string description: Error overview. required: - message ValidationException: description: Validation error content: application/json: schema: type: object properties: message: type: string description: Errors overview. errors: type: object description: A detailed description of each field that failed validation. additionalProperties: type: array items: type: string required: - message - errors securitySchemes: http: type: http scheme: bearer ````