Capabilities catalog
Each capability speaks the same auth, the same envelope, and the same billing. Pick what your agent needs and ship.
Read, scrape, and interact with the live internet.
Live results from the open web.
Run a search across the live web and return ranked results with title, snippet, URL, and freshness.
POST /v1/web/search{
"query": "latest funding rounds in agent infrastructure",
"limit": 5
}{
"results": [
{
"title": "...",
"url": "https://...",
"snippet": "...",
"published_at": "2026-04-30"
}
]
}Any URL to clean Markdown.
Fetch any URL and return clean Markdown, structured data, and key metadata. Handles SPAs and login walls.
POST /v1/web/scrape{
"url": "https://news.ycombinator.com"
}{
"markdown": "# Hacker News\\n...",
"title": "Hacker News",
"links": [
"..."
]
}Full-page agent inside a browser.
Spin up a real Chromium session, navigate, click, type, and extract. Returns transcript, screenshots, and final state.
POST /v1/browser/run{
"task": "Find the cheapest non-stop flight Amsterdam to Tokyo on June 4."
}{
"result": "$612 on JAL",
"steps": 14,
"screenshots": [
"..."
]
}Reader mode for agents.
A leaner web fetch that strips ads, nav, and chrome and returns just the readable body in Markdown.
POST /v1/web/reader{
"url": "https://en.wikipedia.org/wiki/Sonic_boom"
}{
"markdown": "# Sonic boom\\nA sonic boom is..."
}Generate, edit, and understand images.
Text to image, agent-ready.
Generate a high quality image from a prompt. Supports aspect ratio, style presets, and reference images.
POST /v1/image/generate{
"prompt": "a sonic ring on a violet gradient, minimalist",
"aspect": "1:1"
}{
"url": "https://cdn.suprsonic.ai/img/...png"
}Inpaint, outpaint, restyle.
Edit an existing image with a natural language instruction. Supports masking and reference style transfer.
POST /v1/image/edit{
"image_url": "https://...",
"instruction": "remove the background, keep the subject"
}{
"url": "https://cdn.suprsonic.ai/img/...png"
}Caption, classify, and answer questions about an image.
Vision question answering. Returns dense captions, object lists, OCR text, and answers to specific questions.
POST /v1/image/understand{
"image_url": "https://...",
"question": "How many people are in this photo?"
}{
"answer": "Three.",
"caption": "Three people standing near a whiteboard."
}Pixels to text.
Extract text from images and scanned documents with bounding boxes and confidence scores.
POST /v1/vision/ocr{
"image_url": "https://..."
}{
"text": "Invoice 0021\\nAmount due: 1240",
"blocks": [
{
"box": [
10,
20,
200,
60
],
"text": "Invoice 0021"
}
]
}Create video, charts, slides, and other visual artifacts.
Text or image to short video.
Generate a 5 to 10 second video clip from a prompt or an input image. Returns mp4 and a poster frame.
POST /v1/video/generate{
"prompt": "a paper plane flying over a violet city skyline",
"seconds": 6
}{
"url": "https://cdn.suprsonic.ai/video/...mp4",
"poster": "https://..."
}Data to a clean PNG chart.
Take a small dataset and a chart spec, return a PNG suitable for embedding in agent responses.
POST /v1/chart/render{
"type": "bar",
"x": [
"Q1",
"Q2",
"Q3"
],
"y": [
12,
18,
31
],
"title": "Pipeline"
}{
"url": "https://cdn.suprsonic.ai/chart/...png"
}From outline to deck.
Generate a presentation deck from a topic or outline. Returns a downloadable .pptx and per-slide previews.
POST /v1/slides/generate{
"topic": "Q2 board update for an agent infrastructure startup",
"slides": 8
}{
"url": "https://cdn.suprsonic.ai/deck/...pptx",
"slides": 8
}Transcribe, synthesize, and translate speech.
Transcribe any audio.
Transcribe an audio file with timestamps, speaker labels, and language detection.
POST /v1/voice/transcribe{
"audio_url": "https://..."
}{
"text": "Welcome to the show.",
"language": "en",
"segments": [
{
"start": 0,
"end": 2.1,
"speaker": "S1"
}
]
}Studio voice in one call.
Synthesize speech from text with selectable voice, language, and pacing. Returns mp3 or wav.
POST /v1/voice/synthesize{
"text": "Suprsonic gives agents superpowers.",
"voice": "violet"
}{
"url": "https://cdn.suprsonic.ai/voice/...mp3",
"duration": 2.4
}Audio in one language to text in another.
Direct audio-to-text translation across 50+ languages with timestamps.
POST /v1/voice/translate{
"audio_url": "https://...",
"target": "en"
}{
"text": "Good morning everyone.",
"source_language": "ja"
}Parse, extract, and understand PDFs and files.
PDF to clean structured text.
Extract text, tables, and metadata from any PDF. Preserves headings, lists, and table structure.
POST /v1/documents/pdf{
"url": "https://example.com/report.pdf"
}{
"markdown": "# Annual report\\n...",
"tables": [
{
"rows": [
[
"Q1",
"31%"
]
]
}
]
}Any office doc to structured Markdown.
Parse .docx, .xlsx, .pptx, and .csv into Markdown plus a structured schema.
POST /v1/documents/parse{
"url": "https://example.com/contract.docx"
}{
"markdown": "# Master service agreement\\n...",
"structure": {
"sections": 8
}
}Run, generate, and review code in a sandbox.
Run code in a sandbox.
Run Python or Node in an ephemeral sandbox with file I/O. Returns stdout, stderr, and produced files.
POST /v1/code/execute{
"language": "python",
"code": "import json; print(json.dumps({'pi': 3.14}))"
}{
"stdout": "{\"pi\": 3.14}",
"stderr": "",
"exit_code": 0
}Spec to working code.
Generate complete code from a natural language spec, with tests when requested.
POST /v1/code/generate{
"language": "typescript",
"spec": "function that hashes a string with SHA-256"
}{
"code": "import { createHash } from 'crypto';\\nexport function sha256(...)"
}PR review on autopilot.
Review a diff or file and return prioritized findings: bugs, security issues, suggestions.
POST /v1/code/review{
"diff": "@@ -10,4 +10,8 @@\\n+exec(user_input)"
}{
"findings": [
{
"severity": "high",
"line": 11,
"message": "Arbitrary code execution from user input."
}
]
}Translate, summarize, classify, and analyze text.
Text across 100+ languages.
High quality machine translation with optional formality and glossary controls.
POST /v1/language/translate{
"text": "Hello world",
"target": "ja"
}{
"text": "こんにちは世界"
}Long text to short text.
Summarize an article, transcript, or document into bullet points or a TL;DR.
POST /v1/language/summarize{
"text": "...",
"format": "bullets"
}{
"summary": [
"Suprsonic is a unified API for agent capabilities."
]
}Free-form labels for any text.
Zero-shot text classification against a label set you provide. Returns scored labels.
POST /v1/language/classify{
"text": "I cannot log in to my account.",
"labels": [
"billing",
"auth",
"feature_request"
]
}{
"labels": [
{
"name": "auth",
"score": 0.94
}
]
}Names, places, money, dates.
Pull structured entities from raw text. Returns typed spans with offsets and confidence.
POST /v1/language/entities{
"text": "Acme paid Globex $12,400 on March 4."
}{
"entities": [
{
"type": "ORG",
"text": "Acme"
},
{
"type": "MONEY",
"text": "$12,400"
}
]
}Embed, search, query, and structure information.
Text to vector.
Generate dense vector embeddings for any text. Use them for search, clustering, and memory.
POST /v1/data/embed{
"input": "Suprsonic gives agents superpowers."
}{
"vector": [
0.0123,
-0.0456,
"..."
],
"dimensions": 1536
}Better order for retrieval.
Rerank a list of candidate documents against a query. Higher precision than embedding similarity alone.
POST /v1/data/rerank{
"query": "agent api pricing",
"documents": [
"...",
"...",
"..."
]
}{
"ranking": [
{
"index": 2,
"score": 0.91
},
{
"index": 0,
"score": 0.62
}
]
}Hosted similarity search.
Upsert vectors and metadata into a Suprsonic-managed index, then query by vector or text.
POST /v1/data/search{
"index": "support-faq",
"query": "how do I reset my key?"
}{
"matches": [
{
"id": "faq-12",
"score": 0.88,
"text": "To rotate a key..."
}
]
}Send email, SMS, and schedule events.
Transactional email from an agent.
Send a transactional email from a verified Suprsonic sender. Returns the message id.
POST /v1/comm/email{
"to": "user@example.com",
"subject": "Your report is ready",
"body": "..."
}{
"message_id": "msg_01HX...",
"delivered": true
}Short messages to a phone.
Send an SMS from a Suprsonic number to any reachable phone number worldwide.
POST /v1/comm/sms{
"to": "+31612345678",
"body": "Your code is 102938"
}{
"message_id": "sms_01HX..."
}Book a meeting from an agent.
Find a free slot, create an event with attendees, and send invites. Works without OAuth on the user side.
POST /v1/comm/calendar{
"title": "Suprsonic intro",
"attendees": [
"jane@acme.com"
],
"duration": 30
}{
"event_id": "evt_01HX...",
"start": "2026-05-09T15:00:00Z"
}Persistent vector memory for your agents.
Persistent memory per agent.
Save, recall, and forget facts scoped to an agent. Survives across sessions and processes.
POST /v1/memory/recall{
"agent_id": "ax-1",
"query": "user prefers EU region"
}{
"facts": [
{
"fact": "User prefers EU region",
"saved_at": "2026-04-12"
}
]
}Get started
Free tier covers 1,000 credits per month. No card. Most capabilities cost 1 to 5 credits per call.