abliteration.ai is an enterprise LLM provider and governance gateway. This collection covers the full public API surface: the OpenAI-compatible /v1/chat/completions and /v1/responses endpoints, the Anthropic-compatible /v1/messages endpoint, and the Policy Gateway for enterprise policy control and audit.
\nAll three inference endpoints share the same base URL, authentication scheme, model identifier, credit-metering fields, and rate-limit rules. The only differences are the request/response envelope and SDK compatibility.
\nhttps://api.abliteration.ai/v1
\nAll endpoints are served from this base URL. Set the baseUrl collection variable (or the {{baseUrl}} environment variable) to this value to run every request in the collection without editing URLs individually.
\nEvery request must include credentials in the Authorization header as a Bearer token. Three credential types are supported:
\nAPI key — keys begin with ak_. Send as Authorization: Bearer ak_YOUR_KEY. Obtain from the abliteration.ai dashboard. Supports 120 requests per 60-second window.
\nJWT — short-lived token obtained from POST /api/auth/login. Send as Authorization: Bearer . Supports 30 requests per 60-second window.
\nAnonymous free tier — omit the Authorization header and set X-Free-Tier: 1 (for /v1/messages) or X-Free-Tier: true (for /v1/responses and /v1/chat/completions). Limited to 1 free request per device. Does not consume credits.
\nSet the apiKey collection variable to your ak_... key. The collection-level Authorization tab is pre-configured to inject it as a Bearer token on every request.
\nOpenAI-compatible chat completions. Drop-in replacement for OpenAI clients: change base_url to https://api.abliteration.ai/v1 and swap your API key. Supports streaming (stream: true), vision/multimodal content arrays, and the full OpenAI message format. Works with the OpenAI Python SDK, Node.js SDK, LangChain, LlamaIndex, Vercel AI SDK, and any other OpenAI-compatible client.
\nOpenAI Responses API-compatible endpoint. Accepts either a plain string input or a structured Responses-format input array with multimodal parts (input_text, input_image). Supports tool definitions, tool_choice, instructions (system-style), streaming via SSE (response.created, response.output_text.delta, response.completed events), and the flagged_categories moderation field. Use with the OpenAI Python/Node SDK via client.responses.create(), or with Codex CLI by setting base_url to https://api.abliteration.ai/v1.
\nAnthropic Messages API-compatible endpoint. Drop-in replacement for Anthropic SDK clients: set base_url to https://api.abliteration.ai and swap your key. Accepts the Anthropic messages array format (role: user/assistant, content as string or content-block array for multimodal inputs), system prompt, max_tokens, temperature, and stream. Returns the Anthropic response envelope (id, type, role, content blocks, stop_reason, stop_sequence, usage) plus abliteration.ai credit-metering fields. Also supports POST /v1/messages/count_tokens for pre-call budget checks. Use with the Anthropic Python SDK, Claude Code (set ANTHROPIC_BASE_URL + ANTHROPIC_AUTH_TOKEN), or OpenClaw.
\nLimits are enforced per user in a rolling 60-second window. API key callers: 120 req/60s. JWT/UI callers: 30 req/60s. Exceeding the limit returns HTTP 429 with a Retry-After header. The rate limiter is fail-open — over-limit requests are allowed but logged.
\nCredits are billed on total tokens (input + output). Credits per call: ceil(total_tokens / 500), minimum 1 credit. Pricing: ~$5 per 1M tokens. Anonymous free-tier calls do not consume credits. Insufficient credits return HTTP 402. Every response includes remaining_credits, estimated_credits_used, and estimated_cost_usd fields.
\n400 Bad Request — Malformed JSON, missing required fields (model, messages/input), or invalid field values. Check that model is \"abliterated-model\" and the message array is non-empty.
\n401 Unauthorized — Missing or invalid Bearer token. Verify your ak_ key is correct and sent in the Authorization header.
\n402 Insufficient Credits — Credit balance is zero. Purchase credits from the abliteration.ai pricing page or dashboard.
\n429 Rate Limit — Exceeded 120 req/min (API key) or 30 req/min (JWT). Back off exponentially and respect the Retry-After header.
\n502 Bad Gateway — Upstream model temporarily unavailable. Retry after a short delay.
\n504 Gateway Timeout — Model did not respond in time. Try a shorter prompt or retry.
\nDeveloper docs: https://abliteration.ai/docs
\nOpenAI Responses API reference: https://abliteration.ai/docs/openai-responses-api
\nAnthropic Messages API reference: https://abliteration.ai/docs/anthropic-messages-api
\nPolicy Gateway guide: https://abliteration.ai/docs/policy-gateway-backend-guide
\nOpenAPI spec: https://api.abliteration.ai/.well-known/openapi.json
\n","schema":"https://schema.getpostman.com/json/collection/v2.0.0/collection.json","toc":[{"content":"abliteration.ai API","slug":"abliterationai-api"}],"owner":"51220842","collectionId":"4fb7cfce-c445-4495-a6e2-e2d25a61e651","publishedId":"2sBXVcmCop","public":true,"customColor":{"top-bar":"FFFFFF","right-sidebar":"303030","highlight":"FF6C37"},"publishDate":"2026-01-04T03:32:49.000Z"},"item":[{"name":"/chat/completions","id":"340fc381-aa08-4852-830c-e27edbc9229d","protocolProfileBehavior":{"disableBodyPruning":true},"request":{"method":"POST","header":[],"body":{"mode":"raw","raw":"{\n \"model\": \"abliterated-model\",\n \"messages\": [{\"role\":\"user\",\"content\":[\n {\"type\":\"text\",\"text\":\"What’s in this image?\"},\n {\"type\":\"image_url\",\"image_url\":{\"url\":\"https://abliteration.ai/stonehenge.jpg\"}}\n ]}],\n \"stream\": true\n }","options":{"raw":{"language":"json"}}},"url":"https://api.abliteration.ai/v1/chat/completions","description":"This collection provides access to the abliteration.ai Chat Completions API. Integrate directly with uncensored models via API. Full /v1/chat/completions parity for developers who need unrestricted models. Usage-based token pricing with no monthly seats—own your data and keep every request private. This provides vision and text capabilities.
Requires an API key passed in the Authorization header:
Authorization: Bearer YOUR_API_KEY\n\nmodel (string, required)The AI model identifier to use for generating completions.
\n\"abliterated-model\"messages (array, required)An array of message objects representing the conversation history. Each message contains:
\nrole (string): The role of the message sender
\"user\" - Messages from the user
\"assistant\" - Messages from the AI
\"system\" - System instructions
content (array or string): The message content, which can be:
A simple string for text-only messages
\nAn array of content objects for multimodal messages:
\nText content: {\"type\": \"text\", \"text\": \"Your message here\"}
Image content: {\"type\": \"image_url\", \"image_url\": {\"url\": \"https://example.com/image.jpg\"}}
stream (boolean, optional)Controls whether the response is streamed incrementally or returned all at once.
\ntrue - Stream response chunks as they're generated (recommended for real-time UX)
false - Return complete response once generation is finished
Default: false
When stream: true, the API returns Server-Sent Events (SSE) with the following structure:
{\n \"id\": \"chatcmpl-8fece7f7cb930ba2\",\n \"choices\": [{\n \"delta\": {\n \"content\": \"response text\",\n \"role\": \"assistant\",\n \"reasoning\": \"reasoning text\",\n \"reasoning_content\": \"reasoning text\"\n },\n \"finish_reason\": null,\n \"index\": 0\n }],\n \"created\": 1767496111,\n \"model\": \"abliterated-model\",\n \"object\": \"chat.completion.chunk\"\n}\n\nid: Unique identifier for this completion
choices: Array of completion choices (typically one)
delta: The incremental content for this chunk
content: The actual response text (null for reasoning-only chunks)
reasoning: Chain-of-thought reasoning (if available)
role: The role of the responder (usually \"assistant\")
finish_reason: Reason for completion ending (null while streaming, \"stop\" when complete)
created: Unix timestamp of when the completion was created
model: The model used for generation
object: Type of object returned (\"chat.completion.chunk\" for streaming)
Send an image URL along with a text question to get AI-powered image analysis:
\n{\n \"model\": \"abliterated-model\",\n \"messages\": [{\n \"role\": \"user\",\n \"content\": [\n {\n \"type\": \"text\",\n \"text\": \"What's in this image?\"\n },\n {\n \"type\": \"image_url\",\n \"image_url\": {\n \"url\": \"https://abliteration.ai/stonehenge.jpg\"\n }\n }\n ]\n }],\n \"stream\": true\n}\n\nThe AI will analyze the image and provide a detailed description, answering your question about the visual content.
\nImages can be provided as URLs (must be publicly accessible)
\nMultiple images can be included in a single message
\nThe model supports both text-only and multimodal conversations
\nStreaming responses include reasoning tokens that show the model's thought process
\nOpenAI Responses API-compatible endpoint. Existing OpenAI Responses API clients work with a base-URL and API-key switch.
\nBase URL: https://api.abliteration.ai/v1
\nModel ID: abliterated-model
\nmodel (string, required) — Model ID. Use \"abliterated-model\".
\ninput (string | array, optional) — Plain text string or structured Responses-format input array. Use input_text parts for text and input_image parts for image URLs. Video inputs are not supported.
\ninstructions (string, optional) — Optional system-style instruction prepended to the conversation.
\nstream (boolean, optional) — Default false. Set to true to receive Server-Sent Events. Streaming events: response.created, response.output_text.delta, response.completed.
\ntemperature (float, optional) — Sampling temperature 0–2.
\nmax_output_tokens (integer, optional) — Upper bound for generated output tokens.
\ntools (array, optional) — Tool definitions forwarded to the backend Responses implementation.
\ntool_choice (string | object, optional) — Controls whether/how the model may call tools.
\nflagged_categories (array, optional) — Moderation categories to block before inference.
\nReturns a Responses-style object. Key fields: id, object (\"response\"), created_at, status (\"completed\"), model, output (array of message objects with role: \"assistant\" and content blocks of type \"output_text\"), usage (input_tokens, output_tokens, total_tokens), remaining_credits, estimated_credits_used, estimated_cost_usd.
\nAuthorization: Bearer ak_YOUR_KEY. Set the apiKey collection variable to inject automatically. Anonymous free-tier: omit Authorization header and add X-Free-Tier: true header (1 free request per device).
\n120 req/60s (API key), 30 req/60s (JWT). Returns HTTP 429 + Retry-After on limit. Credits: ceil(total_tokens/500), min 1. ~$5/1M tokens. HTTP 402 if balance is zero. Response includes remaining_credits, estimated_credits_used, estimated_cost_usd.
\nFull reference: https://abliteration.ai/docs/openai-responses-api
\n","auth":{"type":"bearer","bearer":{"basicConfig":[{"key":"token","value":"{{ABLITERATION_API_KEY}}"}]},"isInherited":true,"source":{"_postman_id":"4fb7cfce-c445-4495-a6e2-e2d25a61e651","id":"4fb7cfce-c445-4495-a6e2-e2d25a61e651","name":"api.abliteration.ai","type":"collection"}},"urlObject":{"protocol":"https","path":["v1","responses"],"host":["api","abliteration","ai"],"query":[],"variable":[]}},"response":[],"_postman_id":"01dd3c85-06a6-481d-b519-c7750c670edd"},{"name":"/v1/messages","id":"2b3fb467-3e8c-4f48-991a-d6e9be507674","protocolProfileBehavior":{"disableBodyPruning":true},"request":{"method":"POST","header":[],"body":{"mode":"raw","raw":"{\n \"model\": \"abliterated-model\",\n \"max_tokens\": 1024,\n \"system\": \"You are a helpful AI assistant.\",\n \"messages\": [\n {\n \"role\": \"user\",\n \"content\": \"Explain abliteration in one paragraph.\"\n }\n ],\n \"stream\": false,\n \"temperature\": 0.7\n}","options":{"raw":{"language":"json"}}},"url":"https://api.abliteration.ai/v1/messages","description":"Anthropic Messages API-compatible endpoint. Existing Anthropic SDK clients and tools work with a base-URL switch.
\nBase URL: https://api.abliteration.ai/v1
\nModel ID: abliterated-model
\nmodel (string, required) — Model ID. Use \"abliterated-model\".
\nmessages (array, required) — Non-empty array of message objects with role (\"user\" or \"assistant\") and content. Content can be a plain string or an array of content blocks for multimodal inputs (text + images).
\nmax_tokens (integer, optional) — Maximum number of tokens to generate.
\ntemperature (float, optional) — Sampling temperature (0–2).
\nstream (boolean, optional) — Default false. Set to true for SSE streaming.
\nsystem (string | array, optional) — System prompt prepended to the conversation.
\nWhen stream is false (default), the full response is returned as JSON. The response includes credit metering fields specific to abliteration.ai: remaining_credits, estimated_credits_used, and estimated_cost_usd. Key response fields: id, type (\"message\"), role (\"assistant\"), model, content (array of text blocks), stop_reason, stop_sequence, usage (input_tokens, output_tokens).
\nSet stream: true to receive Server-Sent Events (SSE). Each event has an event: line and a data: line with JSON. Events follow the Anthropic streaming protocol: message_start, content_block_start, content_block_delta, content_block_stop, message_delta, and message_stop.
\nPOST /v1/messages/count_tokens returns an estimated input token count without generating a response. Useful for budget checks before sending expensive prompts. Request uses the same schema as /v1/messages. Response: { \"input_tokens\": N }
\nAuthorization: Bearer ak_YOUR_KEY. Set the apiKey collection variable to inject automatically. API keys start with ak_. JWT tokens can be obtained from POST /api/auth/login. Anonymous free-tier: omit Authorization header and add X-Free-Tier: 1 header (1 free request per device, limited throughput).
\n120 req/60s (API key), 30 req/60s (JWT/UI). Returns HTTP 429 + Retry-After on limit. Rate-limiter failures are fail-open (requests allowed, incident logged). Credits: ceil(total_tokens/500), minimum 1 credit per call. Pricing: ~$5/1M tokens. HTTP 402 if credit balance is zero. Response includes remaining_credits, estimated_credits_used, estimated_cost_usd. Anonymous free-tier calls do not consume credits.
\n400 Bad Request: Verify model is present, messages is a non-empty array, and JSON is well-formed.
\n401 Unauthorized: Check API key or JWT is valid and sent as Bearer token in Authorization header.
\n402 Insufficient Credits: Credit balance is zero. Purchase more credits from the pricing page or dashboard.
\n429 Rate Limit: Exceeded 120 req/min (API key) or 30 req/min (UI/JWT). Back off and respect the Retry-After header.
\n502 Bad Gateway: Upstream model is temporarily unavailable. Retry after a short delay.
\n504 Gateway Timeout: Model did not respond in time. Try a shorter prompt or retry.
\nFull reference: https://abliteration.ai/docs/anthropic-messages-api
\n","auth":{"type":"bearer","bearer":{"basicConfig":[{"key":"token","value":"{{ABLITERATION_API_KEY}}"}]},"isInherited":true,"source":{"_postman_id":"4fb7cfce-c445-4495-a6e2-e2d25a61e651","id":"4fb7cfce-c445-4495-a6e2-e2d25a61e651","name":"api.abliteration.ai","type":"collection"}},"urlObject":{"protocol":"https","path":["v1","messages"],"host":["api","abliteration","ai"],"query":[],"variable":[]}},"response":[],"_postman_id":"2b3fb467-3e8c-4f48-991a-d6e9be507674"}],"auth":{"type":"bearer","bearer":{"basicConfig":[{"key":"token","value":"{{ABLITERATION_API_KEY}}"}]}},"event":[{"listen":"prerequest","script":{"id":"c2f819e1-718a-4670-8411-705bf2ce9008","type":"text/javascript","packages":{},"requests":{},"exec":[""]}},{"listen":"test","script":{"id":"de544073-409d-412d-b7c4-c642bf6df993","type":"text/javascript","packages":{},"requests":{},"exec":[""]}}]}